1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.3 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
156 @section Option Summary
158 Here is a summary of all the options, grouped by type. Explanations are
159 in the following sections.
162 @item Overall Options
163 @xref{Overall Options,,Options Controlling the Kind of Output}.
164 @gccoptlist{-c -S -E -o @var{file} -no-canonical-prefixes @gol
165 -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
167 --version -wrapper @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
168 -fdump-ada-spec@r{[}-slim@r{]}} -fdump-go-spec=@var{file}
170 @item C Language Options
171 @xref{C Dialect Options,,Options Controlling C Dialect}.
172 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
173 -aux-info @var{filename} @gol
174 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
175 -fhosted -ffreestanding -fopenmp -fms-extensions -fplan9-extensions @gol
176 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
177 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
178 -fsigned-bitfields -fsigned-char @gol
179 -funsigned-bitfields -funsigned-char}
181 @item C++ Language Options
182 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
183 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
184 -fconserve-space -fconstexpr-depth=@var{n} -ffriend-injection @gol
185 -fno-elide-constructors @gol
186 -fno-enforce-eh-specs @gol
187 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
188 -fno-implicit-templates @gol
189 -fno-implicit-inline-templates @gol
190 -fno-implement-inlines -fms-extensions @gol
191 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
192 -fno-optional-diags -fpermissive @gol
193 -fno-pretty-templates @gol
194 -frepo -fno-rtti -fstats -ftemplate-depth=@var{n} @gol
195 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
196 -fno-default-inline -fvisibility-inlines-hidden @gol
197 -fvisibility-ms-compat @gol
198 -Wabi -Wconversion-null -Wctor-dtor-privacy @gol
199 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
200 -Weffc++ -Wstrict-null-sentinel @gol
201 -Wno-non-template-friend -Wold-style-cast @gol
202 -Woverloaded-virtual -Wno-pmf-conversions @gol
205 @item Objective-C and Objective-C++ Language Options
206 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
207 Objective-C and Objective-C++ Dialects}.
208 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
209 -fgnu-runtime -fnext-runtime @gol
210 -fno-nil-receivers @gol
211 -fobjc-abi-version=@var{n} @gol
212 -fobjc-call-cxx-cdtors @gol
213 -fobjc-direct-dispatch @gol
214 -fobjc-exceptions @gol
217 -fobjc-std=objc1 @gol
218 -freplace-objc-classes @gol
221 -Wassign-intercept @gol
222 -Wno-protocol -Wselector @gol
223 -Wstrict-selector-match @gol
224 -Wundeclared-selector}
226 @item Language Independent Options
227 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
228 @gccoptlist{-fmessage-length=@var{n} @gol
229 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
230 -fno-diagnostics-show-option}
232 @item Warning Options
233 @xref{Warning Options,,Options to Request or Suppress Warnings}.
234 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -pedantic @gol
235 -pedantic-errors @gol
236 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
237 -Wno-attributes -Wno-builtin-macro-redefined @gol
238 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
239 -Wchar-subscripts -Wclobbered -Wcomment @gol
240 -Wconversion -Wcoverage-mismatch -Wcpp -Wno-deprecated @gol
241 -Wno-deprecated-declarations -Wdisabled-optimization @gol
242 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
243 -Wno-endif-labels -Werror -Werror=* @gol
244 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
245 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
246 -Wformat-security -Wformat-y2k @gol
247 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
248 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
249 -Winit-self -Winline @gol
250 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
251 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
252 -Wlogical-op -Wlong-long @gol
253 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
254 -Wmissing-format-attribute -Wmissing-include-dirs @gol
256 -Wno-multichar -Wnonnull -Wno-overflow @gol
257 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
258 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
259 -Wpointer-arith -Wno-pointer-to-int-cast @gol
260 -Wredundant-decls @gol
261 -Wreturn-type -Wsequence-point -Wshadow @gol
262 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
263 -Wstrict-aliasing -Wstrict-aliasing=n @gol
264 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
265 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol
266 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
267 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
268 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
269 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
270 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value @gol
271 -Wunused-variable -Wunused-but-set-parameter -Wunused-but-set-variable @gol
272 -Wvariadic-macros -Wvla -Wvolatile-register-var -Wwrite-strings}
274 @item C and Objective-C-only Warning Options
275 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
276 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
277 -Wold-style-declaration -Wold-style-definition @gol
278 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
279 -Wdeclaration-after-statement -Wpointer-sign}
281 @item Debugging Options
282 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
283 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
284 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
285 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
286 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
287 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
288 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
289 -fdump-statistics @gol
291 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
295 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
299 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
304 -fdump-tree-nrv -fdump-tree-vect @gol
305 -fdump-tree-sink @gol
306 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
307 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
308 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
309 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
310 -ftree-vectorizer-verbose=@var{n} @gol
311 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
312 -fdump-final-insns=@var{file} @gol
313 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
314 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
315 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
316 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
317 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
318 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
319 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
320 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
321 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
322 -ggdb -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
323 -gvms -gxcoff -gxcoff+ @gol
324 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
325 -fdebug-prefix-map=@var{old}=@var{new} @gol
326 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
327 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
328 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
329 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
330 -print-prog-name=@var{program} -print-search-dirs -Q @gol
331 -print-sysroot -print-sysroot-headers-suffix @gol
332 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
334 @item Optimization Options
335 @xref{Optimize Options,,Options that Control Optimization}.
336 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
337 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
338 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
339 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
340 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
341 -fcompare-elim -fcprop-registers -fcrossjumping @gol
342 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
343 -fcx-limited-range @gol
344 -fdata-sections -fdce -fdce -fdelayed-branch @gol
345 -fdelete-null-pointer-checks -fdse -fdevirtualize -fdse @gol
346 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
347 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
348 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
349 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
350 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
351 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
352 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
353 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
354 -fipa-struct-reorg -fira-algorithm=@var{algorithm} @gol
355 -fira-region=@var{region} @gol
356 -fira-loop-pressure -fno-ira-share-save-slots @gol
357 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
358 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
359 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
360 -floop-parallelize-all -flto -flto-compression-level
361 -flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol
362 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
363 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
364 -fno-default-inline @gol
365 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
366 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
367 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
368 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
369 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
370 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
371 -fprefetch-loop-arrays @gol
372 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
373 -fprofile-generate=@var{path} @gol
374 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
375 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
376 -freorder-blocks-and-partition -freorder-functions @gol
377 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
378 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
379 -fsched-spec-load -fsched-spec-load-dangerous @gol
380 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
381 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
382 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
383 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
384 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
385 -fselective-scheduling -fselective-scheduling2 @gol
386 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
387 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
388 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
389 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
391 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
392 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
393 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
394 -ftree-loop-if-convert-stores -ftree-loop-im @gol
395 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
396 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
397 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
398 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
399 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
400 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
401 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
402 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
403 -fwhole-program -fwpa -fuse-linker-plugin @gol
404 --param @var{name}=@var{value}
405 -O -O0 -O1 -O2 -O3 -Os -Ofast}
407 @item Preprocessor Options
408 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
409 @gccoptlist{-A@var{question}=@var{answer} @gol
410 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
411 -C -dD -dI -dM -dN @gol
412 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
413 -idirafter @var{dir} @gol
414 -include @var{file} -imacros @var{file} @gol
415 -iprefix @var{file} -iwithprefix @var{dir} @gol
416 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
417 -imultilib @var{dir} -isysroot @var{dir} @gol
418 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
419 -P -fworking-directory -remap @gol
420 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
421 -Xpreprocessor @var{option}}
423 @item Assembler Option
424 @xref{Assembler Options,,Passing Options to the Assembler}.
425 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
428 @xref{Link Options,,Options for Linking}.
429 @gccoptlist{@var{object-file-name} -l@var{library} @gol
430 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
431 -s -static -static-libgcc -static-libstdc++ -shared @gol
432 -shared-libgcc -symbolic @gol
433 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
436 @item Directory Options
437 @xref{Directory Options,,Options for Directory Search}.
438 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir}}
439 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-
442 @item Machine Dependent Options
443 @xref{Submodel Options,,Hardware Models and Configurations}.
444 @c This list is ordered alphanumerically by subsection name.
445 @c Try and put the significant identifier (CPU or system) first,
446 @c so users have a clue at guessing where the ones they want will be.
449 @gccoptlist{-EB -EL @gol
450 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
451 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
454 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
455 -mabi=@var{name} @gol
456 -mapcs-stack-check -mno-apcs-stack-check @gol
457 -mapcs-float -mno-apcs-float @gol
458 -mapcs-reentrant -mno-apcs-reentrant @gol
459 -msched-prolog -mno-sched-prolog @gol
460 -mlittle-endian -mbig-endian -mwords-little-endian @gol
461 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
462 -mfp16-format=@var{name}
463 -mthumb-interwork -mno-thumb-interwork @gol
464 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
465 -mstructure-size-boundary=@var{n} @gol
466 -mabort-on-noreturn @gol
467 -mlong-calls -mno-long-calls @gol
468 -msingle-pic-base -mno-single-pic-base @gol
469 -mpic-register=@var{reg} @gol
470 -mnop-fun-dllimport @gol
471 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
472 -mpoke-function-name @gol
474 -mtpcs-frame -mtpcs-leaf-frame @gol
475 -mcaller-super-interworking -mcallee-super-interworking @gol
477 -mword-relocations @gol
478 -mfix-cortex-m3-ldrd}
481 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
482 -mcall-prologues -mtiny-stack -mint8}
484 @emph{Blackfin Options}
485 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
486 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
487 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
488 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
489 -mno-id-shared-library -mshared-library-id=@var{n} @gol
490 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
491 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
492 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
496 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
497 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
498 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
499 -mstack-align -mdata-align -mconst-align @gol
500 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
501 -melf -maout -melinux -mlinux -sim -sim2 @gol
502 -mmul-bug-workaround -mno-mul-bug-workaround}
505 @gccoptlist{-mmac -mpush-args}
507 @emph{Darwin Options}
508 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
509 -arch_only -bind_at_load -bundle -bundle_loader @gol
510 -client_name -compatibility_version -current_version @gol
512 -dependency-file -dylib_file -dylinker_install_name @gol
513 -dynamic -dynamiclib -exported_symbols_list @gol
514 -filelist -flat_namespace -force_cpusubtype_ALL @gol
515 -force_flat_namespace -headerpad_max_install_names @gol
517 -image_base -init -install_name -keep_private_externs @gol
518 -multi_module -multiply_defined -multiply_defined_unused @gol
519 -noall_load -no_dead_strip_inits_and_terms @gol
520 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
521 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
522 -private_bundle -read_only_relocs -sectalign @gol
523 -sectobjectsymbols -whyload -seg1addr @gol
524 -sectcreate -sectobjectsymbols -sectorder @gol
525 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
526 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
527 -segprot -segs_read_only_addr -segs_read_write_addr @gol
528 -single_module -static -sub_library -sub_umbrella @gol
529 -twolevel_namespace -umbrella -undefined @gol
530 -unexported_symbols_list -weak_reference_mismatches @gol
531 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
532 -mkernel -mone-byte-bool}
534 @emph{DEC Alpha Options}
535 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
536 -mieee -mieee-with-inexact -mieee-conformant @gol
537 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
538 -mtrap-precision=@var{mode} -mbuild-constants @gol
539 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
540 -mbwx -mmax -mfix -mcix @gol
541 -mfloat-vax -mfloat-ieee @gol
542 -mexplicit-relocs -msmall-data -mlarge-data @gol
543 -msmall-text -mlarge-text @gol
544 -mmemory-latency=@var{time}}
546 @emph{DEC Alpha/VMS Options}
547 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
550 @gccoptlist{-msmall-model -mno-lsim}
553 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
554 -mhard-float -msoft-float @gol
555 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
556 -mdouble -mno-double @gol
557 -mmedia -mno-media -mmuladd -mno-muladd @gol
558 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
559 -mlinked-fp -mlong-calls -malign-labels @gol
560 -mlibrary-pic -macc-4 -macc-8 @gol
561 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
562 -moptimize-membar -mno-optimize-membar @gol
563 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
564 -mvliw-branch -mno-vliw-branch @gol
565 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
566 -mno-nested-cond-exec -mtomcat-stats @gol
570 @emph{GNU/Linux Options}
571 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
572 -tno-android-cc -tno-android-ld}
574 @emph{H8/300 Options}
575 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
578 @gccoptlist{-march=@var{architecture-type} @gol
579 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
580 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
581 -mfixed-range=@var{register-range} @gol
582 -mjump-in-delay -mlinker-opt -mlong-calls @gol
583 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
584 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
585 -mno-jump-in-delay -mno-long-load-store @gol
586 -mno-portable-runtime -mno-soft-float @gol
587 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
588 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
589 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
590 -munix=@var{unix-std} -nolibdld -static -threads}
592 @emph{i386 and x86-64 Options}
593 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
594 -mfpmath=@var{unit} @gol
595 -masm=@var{dialect} -mno-fancy-math-387 @gol
596 -mno-fp-ret-in-387 -msoft-float @gol
597 -mno-wide-multiply -mrtd -malign-double @gol
598 -mpreferred-stack-boundary=@var{num}
599 -mincoming-stack-boundary=@var{num} @gol
600 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip -mvzeroupper @gol
601 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
602 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfused-madd @gol
603 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlwp @gol
604 -mthreads -mno-align-stringops -minline-all-stringops @gol
605 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
606 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
607 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
608 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
609 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
610 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
611 -mcmodel=@var{code-model} -mabi=@var{name} @gol
612 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
613 -msse2avx -mfentry -m8bit-idiv}
616 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
617 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
618 -mconstant-gp -mauto-pic -mfused-madd @gol
619 -minline-float-divide-min-latency @gol
620 -minline-float-divide-max-throughput @gol
621 -mno-inline-float-divide @gol
622 -minline-int-divide-min-latency @gol
623 -minline-int-divide-max-throughput @gol
624 -mno-inline-int-divide @gol
625 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
626 -mno-inline-sqrt @gol
627 -mdwarf2-asm -mearly-stop-bits @gol
628 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
629 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
630 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
631 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
632 -msched-spec-ldc -msched-spec-control-ldc @gol
633 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
634 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
635 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
636 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
638 @emph{IA-64/VMS Options}
639 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
642 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
643 -msign-extend-enabled -muser-enabled}
645 @emph{M32R/D Options}
646 @gccoptlist{-m32r2 -m32rx -m32r @gol
648 -malign-loops -mno-align-loops @gol
649 -missue-rate=@var{number} @gol
650 -mbranch-cost=@var{number} @gol
651 -mmodel=@var{code-size-model-type} @gol
652 -msdata=@var{sdata-type} @gol
653 -mno-flush-func -mflush-func=@var{name} @gol
654 -mno-flush-trap -mflush-trap=@var{number} @gol
658 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
660 @emph{M680x0 Options}
661 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
662 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
663 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
664 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
665 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
666 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
667 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
668 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
671 @emph{M68hc1x Options}
672 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
673 -mauto-incdec -minmax -mlong-calls -mshort @gol
674 -msoft-reg-count=@var{count}}
677 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
678 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
679 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
680 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
681 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
684 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
685 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
686 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
687 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
690 @emph{MicroBlaze Options}
691 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
692 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
693 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
694 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
695 -mxl-mode-@var{app-model}}
698 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
699 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
700 -mips64 -mips64r2 @gol
701 -mips16 -mno-mips16 -mflip-mips16 @gol
702 -minterlink-mips16 -mno-interlink-mips16 @gol
703 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
704 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
705 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
706 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
707 -mfpu=@var{fpu-type} @gol
708 -msmartmips -mno-smartmips @gol
709 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
710 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
711 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
712 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
713 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
714 -membedded-data -mno-embedded-data @gol
715 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
716 -mcode-readable=@var{setting} @gol
717 -msplit-addresses -mno-split-addresses @gol
718 -mexplicit-relocs -mno-explicit-relocs @gol
719 -mcheck-zero-division -mno-check-zero-division @gol
720 -mdivide-traps -mdivide-breaks @gol
721 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
722 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
723 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
724 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
725 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
726 -mflush-func=@var{func} -mno-flush-func @gol
727 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
728 -mfp-exceptions -mno-fp-exceptions @gol
729 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
730 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
733 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
734 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
735 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
736 -mno-base-addresses -msingle-exit -mno-single-exit}
738 @emph{MN10300 Options}
739 @gccoptlist{-mmult-bug -mno-mult-bug @gol
740 -mno-am33 -mam33 -mam33-2 -mam34 @gol
741 -mtune=@var{cpu-type} @gol
742 -mreturn-pointer-on-d0 @gol
743 -mno-crt0 -mrelax -mliw}
745 @emph{PDP-11 Options}
746 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
747 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
748 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
749 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
750 -mbranch-expensive -mbranch-cheap @gol
751 -munix-asm -mdec-asm}
753 @emph{picoChip Options}
754 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
755 -msymbol-as-address -mno-inefficient-warnings}
757 @emph{PowerPC Options}
758 See RS/6000 and PowerPC Options.
760 @emph{RS/6000 and PowerPC Options}
761 @gccoptlist{-mcpu=@var{cpu-type} @gol
762 -mtune=@var{cpu-type} @gol
763 -mcmodel=@var{code-model} @gol
764 -mpower -mno-power -mpower2 -mno-power2 @gol
765 -mpowerpc -mpowerpc64 -mno-powerpc @gol
766 -maltivec -mno-altivec @gol
767 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
768 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
769 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
770 -mfprnd -mno-fprnd @gol
771 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
772 -mnew-mnemonics -mold-mnemonics @gol
773 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
774 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
775 -malign-power -malign-natural @gol
776 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
777 -msingle-float -mdouble-float -msimple-fpu @gol
778 -mstring -mno-string -mupdate -mno-update @gol
779 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
780 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
781 -mstrict-align -mno-strict-align -mrelocatable @gol
782 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
783 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
784 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
785 -mprioritize-restricted-insns=@var{priority} @gol
786 -msched-costly-dep=@var{dependence_type} @gol
787 -minsert-sched-nops=@var{scheme} @gol
788 -mcall-sysv -mcall-netbsd @gol
789 -maix-struct-return -msvr4-struct-return @gol
790 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
791 -mblock-move-inline-limit=@var{num} @gol
792 -misel -mno-isel @gol
793 -misel=yes -misel=no @gol
795 -mspe=yes -mspe=no @gol
797 -mgen-cell-microcode -mwarn-cell-microcode @gol
798 -mvrsave -mno-vrsave @gol
799 -mmulhw -mno-mulhw @gol
800 -mdlmzb -mno-dlmzb @gol
801 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
802 -mprototype -mno-prototype @gol
803 -msim -mmvme -mads -myellowknife -memb -msdata @gol
804 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
805 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision
806 -mno-recip-precision @gol
807 -mveclibabi=@var{type} -mfriz -mno-friz}
810 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
812 -mbig-endian-data -mlittle-endian-data @gol
815 -mas100-syntax -mno-as100-syntax@gol
817 -mmax-constant-size=@gol
819 -msave-acc-in-interrupts}
821 @emph{S/390 and zSeries Options}
822 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
823 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
824 -mlong-double-64 -mlong-double-128 @gol
825 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
826 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
827 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
828 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
829 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
832 @gccoptlist{-meb -mel @gol
836 -mscore5 -mscore5u -mscore7 -mscore7d}
839 @gccoptlist{-m1 -m2 -m2e @gol
840 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
842 -m4-nofpu -m4-single-only -m4-single -m4 @gol
843 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
844 -m5-64media -m5-64media-nofpu @gol
845 -m5-32media -m5-32media-nofpu @gol
846 -m5-compact -m5-compact-nofpu @gol
847 -mb -ml -mdalign -mrelax @gol
848 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
849 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
850 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
851 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
852 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
853 -maccumulate-outgoing-args -minvalid-symbols}
855 @emph{Solaris 2 Options}
856 @gccoptlist{-mimpure-text -mno-impure-text @gol
857 -threads -pthreads -pthread}
860 @gccoptlist{-mcpu=@var{cpu-type} @gol
861 -mtune=@var{cpu-type} @gol
862 -mcmodel=@var{code-model} @gol
863 -m32 -m64 -mapp-regs -mno-app-regs @gol
864 -mfaster-structs -mno-faster-structs @gol
865 -mfpu -mno-fpu -mhard-float -msoft-float @gol
866 -mhard-quad-float -msoft-quad-float @gol
868 -mstack-bias -mno-stack-bias @gol
869 -munaligned-doubles -mno-unaligned-doubles @gol
870 -mv8plus -mno-v8plus -mvis -mno-vis}
873 @gccoptlist{-mwarn-reloc -merror-reloc @gol
874 -msafe-dma -munsafe-dma @gol
876 -msmall-mem -mlarge-mem -mstdmain @gol
877 -mfixed-range=@var{register-range} @gol
879 -maddress-space-conversion -mno-address-space-conversion @gol
880 -mcache-size=@var{cache-size} @gol
881 -matomic-updates -mno-atomic-updates}
883 @emph{System V Options}
884 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
887 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
888 -mprolog-function -mno-prolog-function -mspace @gol
889 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
890 -mapp-regs -mno-app-regs @gol
891 -mdisable-callt -mno-disable-callt @gol
894 -mv850e1 -mv850es @gol
899 @gccoptlist{-mg -mgnu -munix}
901 @emph{VxWorks Options}
902 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
903 -Xbind-lazy -Xbind-now}
905 @emph{x86-64 Options}
906 See i386 and x86-64 Options.
908 @emph{i386 and x86-64 Windows Options}
909 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
910 -mnop-fun-dllimport -mthread @gol
911 -municode -mwin32 -mwindows -fno-set-stack-executable}
913 @emph{Xstormy16 Options}
916 @emph{Xtensa Options}
917 @gccoptlist{-mconst16 -mno-const16 @gol
918 -mfused-madd -mno-fused-madd @gol
920 -mserialize-volatile -mno-serialize-volatile @gol
921 -mtext-section-literals -mno-text-section-literals @gol
922 -mtarget-align -mno-target-align @gol
923 -mlongcalls -mno-longcalls}
925 @emph{zSeries Options}
926 See S/390 and zSeries Options.
928 @item Code Generation Options
929 @xref{Code Gen Options,,Options for Code Generation Conventions}.
930 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
931 -ffixed-@var{reg} -fexceptions @gol
932 -fnon-call-exceptions -funwind-tables @gol
933 -fasynchronous-unwind-tables @gol
934 -finhibit-size-directive -finstrument-functions @gol
935 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
936 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
937 -fno-common -fno-ident @gol
938 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
939 -fno-jump-tables @gol
940 -frecord-gcc-switches @gol
941 -freg-struct-return -fshort-enums @gol
942 -fshort-double -fshort-wchar @gol
943 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
944 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
945 -fno-stack-limit -fsplit-stack @gol
946 -fleading-underscore -ftls-model=@var{model} @gol
947 -ftrapv -fwrapv -fbounds-check @gol
948 -fvisibility -fstrict-volatile-bitfields}
952 * Overall Options:: Controlling the kind of output:
953 an executable, object files, assembler files,
954 or preprocessed source.
955 * C Dialect Options:: Controlling the variant of C language compiled.
956 * C++ Dialect Options:: Variations on C++.
957 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
959 * Language Independent Options:: Controlling how diagnostics should be
961 * Warning Options:: How picky should the compiler be?
962 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
963 * Optimize Options:: How much optimization?
964 * Preprocessor Options:: Controlling header files and macro definitions.
965 Also, getting dependency information for Make.
966 * Assembler Options:: Passing options to the assembler.
967 * Link Options:: Specifying libraries and so on.
968 * Directory Options:: Where to find header files and libraries.
969 Where to find the compiler executable files.
970 * Spec Files:: How to pass switches to sub-processes.
971 * Target Options:: Running a cross-compiler, or an old version of GCC.
974 @node Overall Options
975 @section Options Controlling the Kind of Output
977 Compilation can involve up to four stages: preprocessing, compilation
978 proper, assembly and linking, always in that order. GCC is capable of
979 preprocessing and compiling several files either into several
980 assembler input files, or into one assembler input file; then each
981 assembler input file produces an object file, and linking combines all
982 the object files (those newly compiled, and those specified as input)
983 into an executable file.
985 @cindex file name suffix
986 For any given input file, the file name suffix determines what kind of
991 C source code which must be preprocessed.
994 C source code which should not be preprocessed.
997 C++ source code which should not be preprocessed.
1000 Objective-C source code. Note that you must link with the @file{libobjc}
1001 library to make an Objective-C program work.
1004 Objective-C source code which should not be preprocessed.
1008 Objective-C++ source code. Note that you must link with the @file{libobjc}
1009 library to make an Objective-C++ program work. Note that @samp{.M} refers
1010 to a literal capital M@.
1012 @item @var{file}.mii
1013 Objective-C++ source code which should not be preprocessed.
1016 C, C++, Objective-C or Objective-C++ header file to be turned into a
1017 precompiled header (default), or C, C++ header file to be turned into an
1018 Ada spec (via the @option{-fdump-ada-spec} switch).
1021 @itemx @var{file}.cp
1022 @itemx @var{file}.cxx
1023 @itemx @var{file}.cpp
1024 @itemx @var{file}.CPP
1025 @itemx @var{file}.c++
1027 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1028 the last two letters must both be literally @samp{x}. Likewise,
1029 @samp{.C} refers to a literal capital C@.
1033 Objective-C++ source code which must be preprocessed.
1035 @item @var{file}.mii
1036 Objective-C++ source code which should not be preprocessed.
1040 @itemx @var{file}.hp
1041 @itemx @var{file}.hxx
1042 @itemx @var{file}.hpp
1043 @itemx @var{file}.HPP
1044 @itemx @var{file}.h++
1045 @itemx @var{file}.tcc
1046 C++ header file to be turned into a precompiled header or Ada spec.
1049 @itemx @var{file}.for
1050 @itemx @var{file}.ftn
1051 Fixed form Fortran source code which should not be preprocessed.
1054 @itemx @var{file}.FOR
1055 @itemx @var{file}.fpp
1056 @itemx @var{file}.FPP
1057 @itemx @var{file}.FTN
1058 Fixed form Fortran source code which must be preprocessed (with the traditional
1061 @item @var{file}.f90
1062 @itemx @var{file}.f95
1063 @itemx @var{file}.f03
1064 @itemx @var{file}.f08
1065 Free form Fortran source code which should not be preprocessed.
1067 @item @var{file}.F90
1068 @itemx @var{file}.F95
1069 @itemx @var{file}.F03
1070 @itemx @var{file}.F08
1071 Free form Fortran source code which must be preprocessed (with the
1072 traditional preprocessor).
1077 @c FIXME: Descriptions of Java file types.
1083 @item @var{file}.ads
1084 Ada source code file which contains a library unit declaration (a
1085 declaration of a package, subprogram, or generic, or a generic
1086 instantiation), or a library unit renaming declaration (a package,
1087 generic, or subprogram renaming declaration). Such files are also
1090 @item @var{file}.adb
1091 Ada source code file containing a library unit body (a subprogram or
1092 package body). Such files are also called @dfn{bodies}.
1094 @c GCC also knows about some suffixes for languages not yet included:
1105 @itemx @var{file}.sx
1106 Assembler code which must be preprocessed.
1109 An object file to be fed straight into linking.
1110 Any file name with no recognized suffix is treated this way.
1114 You can specify the input language explicitly with the @option{-x} option:
1117 @item -x @var{language}
1118 Specify explicitly the @var{language} for the following input files
1119 (rather than letting the compiler choose a default based on the file
1120 name suffix). This option applies to all following input files until
1121 the next @option{-x} option. Possible values for @var{language} are:
1123 c c-header cpp-output
1124 c++ c++-header c++-cpp-output
1125 objective-c objective-c-header objective-c-cpp-output
1126 objective-c++ objective-c++-header objective-c++-cpp-output
1127 assembler assembler-with-cpp
1129 f77 f77-cpp-input f95 f95-cpp-input
1135 Turn off any specification of a language, so that subsequent files are
1136 handled according to their file name suffixes (as they are if @option{-x}
1137 has not been used at all).
1139 @item -pass-exit-codes
1140 @opindex pass-exit-codes
1141 Normally the @command{gcc} program will exit with the code of 1 if any
1142 phase of the compiler returns a non-success return code. If you specify
1143 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1144 numerically highest error produced by any phase that returned an error
1145 indication. The C, C++, and Fortran frontends return 4, if an internal
1146 compiler error is encountered.
1149 If you only want some of the stages of compilation, you can use
1150 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1151 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1152 @command{gcc} is to stop. Note that some combinations (for example,
1153 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1158 Compile or assemble the source files, but do not link. The linking
1159 stage simply is not done. The ultimate output is in the form of an
1160 object file for each source file.
1162 By default, the object file name for a source file is made by replacing
1163 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1165 Unrecognized input files, not requiring compilation or assembly, are
1170 Stop after the stage of compilation proper; do not assemble. The output
1171 is in the form of an assembler code file for each non-assembler input
1174 By default, the assembler file name for a source file is made by
1175 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1177 Input files that don't require compilation are ignored.
1181 Stop after the preprocessing stage; do not run the compiler proper. The
1182 output is in the form of preprocessed source code, which is sent to the
1185 Input files which don't require preprocessing are ignored.
1187 @cindex output file option
1190 Place output in file @var{file}. This applies regardless to whatever
1191 sort of output is being produced, whether it be an executable file,
1192 an object file, an assembler file or preprocessed C code.
1194 If @option{-o} is not specified, the default is to put an executable
1195 file in @file{a.out}, the object file for
1196 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1197 assembler file in @file{@var{source}.s}, a precompiled header file in
1198 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1203 Print (on standard error output) the commands executed to run the stages
1204 of compilation. Also print the version number of the compiler driver
1205 program and of the preprocessor and the compiler proper.
1209 Like @option{-v} except the commands are not executed and arguments
1210 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1211 This is useful for shell scripts to capture the driver-generated command lines.
1215 Use pipes rather than temporary files for communication between the
1216 various stages of compilation. This fails to work on some systems where
1217 the assembler is unable to read from a pipe; but the GNU assembler has
1222 Print (on the standard output) a description of the command line options
1223 understood by @command{gcc}. If the @option{-v} option is also specified
1224 then @option{--help} will also be passed on to the various processes
1225 invoked by @command{gcc}, so that they can display the command line options
1226 they accept. If the @option{-Wextra} option has also been specified
1227 (prior to the @option{--help} option), then command line options which
1228 have no documentation associated with them will also be displayed.
1231 @opindex target-help
1232 Print (on the standard output) a description of target-specific command
1233 line options for each tool. For some targets extra target-specific
1234 information may also be printed.
1236 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1237 Print (on the standard output) a description of the command line
1238 options understood by the compiler that fit into all specified classes
1239 and qualifiers. These are the supported classes:
1242 @item @samp{optimizers}
1243 This will display all of the optimization options supported by the
1246 @item @samp{warnings}
1247 This will display all of the options controlling warning messages
1248 produced by the compiler.
1251 This will display target-specific options. Unlike the
1252 @option{--target-help} option however, target-specific options of the
1253 linker and assembler will not be displayed. This is because those
1254 tools do not currently support the extended @option{--help=} syntax.
1257 This will display the values recognized by the @option{--param}
1260 @item @var{language}
1261 This will display the options supported for @var{language}, where
1262 @var{language} is the name of one of the languages supported in this
1266 This will display the options that are common to all languages.
1269 These are the supported qualifiers:
1272 @item @samp{undocumented}
1273 Display only those options which are undocumented.
1276 Display options which take an argument that appears after an equal
1277 sign in the same continuous piece of text, such as:
1278 @samp{--help=target}.
1280 @item @samp{separate}
1281 Display options which take an argument that appears as a separate word
1282 following the original option, such as: @samp{-o output-file}.
1285 Thus for example to display all the undocumented target-specific
1286 switches supported by the compiler the following can be used:
1289 --help=target,undocumented
1292 The sense of a qualifier can be inverted by prefixing it with the
1293 @samp{^} character, so for example to display all binary warning
1294 options (i.e., ones that are either on or off and that do not take an
1295 argument), which have a description the following can be used:
1298 --help=warnings,^joined,^undocumented
1301 The argument to @option{--help=} should not consist solely of inverted
1304 Combining several classes is possible, although this usually
1305 restricts the output by so much that there is nothing to display. One
1306 case where it does work however is when one of the classes is
1307 @var{target}. So for example to display all the target-specific
1308 optimization options the following can be used:
1311 --help=target,optimizers
1314 The @option{--help=} option can be repeated on the command line. Each
1315 successive use will display its requested class of options, skipping
1316 those that have already been displayed.
1318 If the @option{-Q} option appears on the command line before the
1319 @option{--help=} option, then the descriptive text displayed by
1320 @option{--help=} is changed. Instead of describing the displayed
1321 options, an indication is given as to whether the option is enabled,
1322 disabled or set to a specific value (assuming that the compiler
1323 knows this at the point where the @option{--help=} option is used).
1325 Here is a truncated example from the ARM port of @command{gcc}:
1328 % gcc -Q -mabi=2 --help=target -c
1329 The following options are target specific:
1331 -mabort-on-noreturn [disabled]
1335 The output is sensitive to the effects of previous command line
1336 options, so for example it is possible to find out which optimizations
1337 are enabled at @option{-O2} by using:
1340 -Q -O2 --help=optimizers
1343 Alternatively you can discover which binary optimizations are enabled
1344 by @option{-O3} by using:
1347 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1348 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1349 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1352 @item -no-canonical-prefixes
1353 @opindex no-canonical-prefixes
1354 Do not expand any symbolic links, resolve references to @samp{/../}
1355 or @samp{/./}, or make the path absolute when generating a relative
1360 Display the version number and copyrights of the invoked GCC@.
1364 Invoke all subcommands under a wrapper program. The name of the
1365 wrapper program and its parameters are passed as a comma separated
1369 gcc -c t.c -wrapper gdb,--args
1372 This will invoke all subprograms of @command{gcc} under
1373 @samp{gdb --args}, thus the invocation of @command{cc1} will be
1374 @samp{gdb --args cc1 @dots{}}.
1376 @item -fplugin=@var{name}.so
1377 Load the plugin code in file @var{name}.so, assumed to be a
1378 shared object to be dlopen'd by the compiler. The base name of
1379 the shared object file is used to identify the plugin for the
1380 purposes of argument parsing (See
1381 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1382 Each plugin should define the callback functions specified in the
1385 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1386 Define an argument called @var{key} with a value of @var{value}
1387 for the plugin called @var{name}.
1389 @item -fdump-ada-spec@r{[}-slim@r{]}
1390 For C and C++ source and include files, generate corresponding Ada
1391 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1392 GNAT User's Guide}, which provides detailed documentation on this feature.
1394 @item -fdump-go-spec=@var{file}
1395 For input files in any language, generate corresponding Go
1396 declarations in @var{file}. This generates Go @code{const},
1397 @code{type}, @code{var}, and @code{func} declarations which may be a
1398 useful way to start writing a Go interface to code written in some
1401 @include @value{srcdir}/../libiberty/at-file.texi
1405 @section Compiling C++ Programs
1407 @cindex suffixes for C++ source
1408 @cindex C++ source file suffixes
1409 C++ source files conventionally use one of the suffixes @samp{.C},
1410 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1411 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1412 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1413 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1414 files with these names and compiles them as C++ programs even if you
1415 call the compiler the same way as for compiling C programs (usually
1416 with the name @command{gcc}).
1420 However, the use of @command{gcc} does not add the C++ library.
1421 @command{g++} is a program that calls GCC and treats @samp{.c},
1422 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1423 files unless @option{-x} is used, and automatically specifies linking
1424 against the C++ library. This program is also useful when
1425 precompiling a C header file with a @samp{.h} extension for use in C++
1426 compilations. On many systems, @command{g++} is also installed with
1427 the name @command{c++}.
1429 @cindex invoking @command{g++}
1430 When you compile C++ programs, you may specify many of the same
1431 command-line options that you use for compiling programs in any
1432 language; or command-line options meaningful for C and related
1433 languages; or options that are meaningful only for C++ programs.
1434 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1435 explanations of options for languages related to C@.
1436 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1437 explanations of options that are meaningful only for C++ programs.
1439 @node C Dialect Options
1440 @section Options Controlling C Dialect
1441 @cindex dialect options
1442 @cindex language dialect options
1443 @cindex options, dialect
1445 The following options control the dialect of C (or languages derived
1446 from C, such as C++, Objective-C and Objective-C++) that the compiler
1450 @cindex ANSI support
1454 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1455 equivalent to @samp{-std=c++98}.
1457 This turns off certain features of GCC that are incompatible with ISO
1458 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1459 such as the @code{asm} and @code{typeof} keywords, and
1460 predefined macros such as @code{unix} and @code{vax} that identify the
1461 type of system you are using. It also enables the undesirable and
1462 rarely used ISO trigraph feature. For the C compiler,
1463 it disables recognition of C++ style @samp{//} comments as well as
1464 the @code{inline} keyword.
1466 The alternate keywords @code{__asm__}, @code{__extension__},
1467 @code{__inline__} and @code{__typeof__} continue to work despite
1468 @option{-ansi}. You would not want to use them in an ISO C program, of
1469 course, but it is useful to put them in header files that might be included
1470 in compilations done with @option{-ansi}. Alternate predefined macros
1471 such as @code{__unix__} and @code{__vax__} are also available, with or
1472 without @option{-ansi}.
1474 The @option{-ansi} option does not cause non-ISO programs to be
1475 rejected gratuitously. For that, @option{-pedantic} is required in
1476 addition to @option{-ansi}. @xref{Warning Options}.
1478 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1479 option is used. Some header files may notice this macro and refrain
1480 from declaring certain functions or defining certain macros that the
1481 ISO standard doesn't call for; this is to avoid interfering with any
1482 programs that might use these names for other things.
1484 Functions that would normally be built in but do not have semantics
1485 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1486 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1487 built-in functions provided by GCC}, for details of the functions
1492 Determine the language standard. @xref{Standards,,Language Standards
1493 Supported by GCC}, for details of these standard versions. This option
1494 is currently only supported when compiling C or C++.
1496 The compiler can accept several base standards, such as @samp{c90} or
1497 @samp{c++98}, and GNU dialects of those standards, such as
1498 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1499 compiler will accept all programs following that standard and those
1500 using GNU extensions that do not contradict it. For example,
1501 @samp{-std=c90} turns off certain features of GCC that are
1502 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1503 keywords, but not other GNU extensions that do not have a meaning in
1504 ISO C90, such as omitting the middle term of a @code{?:}
1505 expression. On the other hand, by specifying a GNU dialect of a
1506 standard, all features the compiler support are enabled, even when
1507 those features change the meaning of the base standard and some
1508 strict-conforming programs may be rejected. The particular standard
1509 is used by @option{-pedantic} to identify which features are GNU
1510 extensions given that version of the standard. For example
1511 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1512 comments, while @samp{-std=gnu99 -pedantic} would not.
1514 A value for this option must be provided; possible values are
1520 Support all ISO C90 programs (certain GNU extensions that conflict
1521 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1523 @item iso9899:199409
1524 ISO C90 as modified in amendment 1.
1530 ISO C99. Note that this standard is not yet fully supported; see
1531 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1532 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1535 ISO C1X, the draft of the next revision of the ISO C standard.
1536 Support is limited and experimental and features enabled by this
1537 option may be changed or removed if changed in or removed from the
1542 GNU dialect of ISO C90 (including some C99 features). This
1543 is the default for C code.
1547 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1548 this will become the default. The name @samp{gnu9x} is deprecated.
1551 GNU dialect of ISO C1X. Support is limited and experimental and
1552 features enabled by this option may be changed or removed if changed
1553 in or removed from the standard draft.
1556 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1560 GNU dialect of @option{-std=c++98}. This is the default for
1564 The working draft of the upcoming ISO C++0x standard. This option
1565 enables experimental features that are likely to be included in
1566 C++0x. The working draft is constantly changing, and any feature that is
1567 enabled by this flag may be removed from future versions of GCC if it is
1568 not part of the C++0x standard.
1571 GNU dialect of @option{-std=c++0x}. This option enables
1572 experimental features that may be removed in future versions of GCC.
1575 @item -fgnu89-inline
1576 @opindex fgnu89-inline
1577 The option @option{-fgnu89-inline} tells GCC to use the traditional
1578 GNU semantics for @code{inline} functions when in C99 mode.
1579 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1580 is accepted and ignored by GCC versions 4.1.3 up to but not including
1581 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1582 C99 mode. Using this option is roughly equivalent to adding the
1583 @code{gnu_inline} function attribute to all inline functions
1584 (@pxref{Function Attributes}).
1586 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1587 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1588 specifies the default behavior). This option was first supported in
1589 GCC 4.3. This option is not supported in @option{-std=c90} or
1590 @option{-std=gnu90} mode.
1592 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1593 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1594 in effect for @code{inline} functions. @xref{Common Predefined
1595 Macros,,,cpp,The C Preprocessor}.
1597 @item -aux-info @var{filename}
1599 Output to the given filename prototyped declarations for all functions
1600 declared and/or defined in a translation unit, including those in header
1601 files. This option is silently ignored in any language other than C@.
1603 Besides declarations, the file indicates, in comments, the origin of
1604 each declaration (source file and line), whether the declaration was
1605 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1606 @samp{O} for old, respectively, in the first character after the line
1607 number and the colon), and whether it came from a declaration or a
1608 definition (@samp{C} or @samp{F}, respectively, in the following
1609 character). In the case of function definitions, a K&R-style list of
1610 arguments followed by their declarations is also provided, inside
1611 comments, after the declaration.
1615 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1616 keyword, so that code can use these words as identifiers. You can use
1617 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1618 instead. @option{-ansi} implies @option{-fno-asm}.
1620 In C++, this switch only affects the @code{typeof} keyword, since
1621 @code{asm} and @code{inline} are standard keywords. You may want to
1622 use the @option{-fno-gnu-keywords} flag instead, which has the same
1623 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1624 switch only affects the @code{asm} and @code{typeof} keywords, since
1625 @code{inline} is a standard keyword in ISO C99.
1628 @itemx -fno-builtin-@var{function}
1629 @opindex fno-builtin
1630 @cindex built-in functions
1631 Don't recognize built-in functions that do not begin with
1632 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1633 functions provided by GCC}, for details of the functions affected,
1634 including those which are not built-in functions when @option{-ansi} or
1635 @option{-std} options for strict ISO C conformance are used because they
1636 do not have an ISO standard meaning.
1638 GCC normally generates special code to handle certain built-in functions
1639 more efficiently; for instance, calls to @code{alloca} may become single
1640 instructions that adjust the stack directly, and calls to @code{memcpy}
1641 may become inline copy loops. The resulting code is often both smaller
1642 and faster, but since the function calls no longer appear as such, you
1643 cannot set a breakpoint on those calls, nor can you change the behavior
1644 of the functions by linking with a different library. In addition,
1645 when a function is recognized as a built-in function, GCC may use
1646 information about that function to warn about problems with calls to
1647 that function, or to generate more efficient code, even if the
1648 resulting code still contains calls to that function. For example,
1649 warnings are given with @option{-Wformat} for bad calls to
1650 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1651 known not to modify global memory.
1653 With the @option{-fno-builtin-@var{function}} option
1654 only the built-in function @var{function} is
1655 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1656 function is named that is not built-in in this version of GCC, this
1657 option is ignored. There is no corresponding
1658 @option{-fbuiltin-@var{function}} option; if you wish to enable
1659 built-in functions selectively when using @option{-fno-builtin} or
1660 @option{-ffreestanding}, you may define macros such as:
1663 #define abs(n) __builtin_abs ((n))
1664 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1669 @cindex hosted environment
1671 Assert that compilation takes place in a hosted environment. This implies
1672 @option{-fbuiltin}. A hosted environment is one in which the
1673 entire standard library is available, and in which @code{main} has a return
1674 type of @code{int}. Examples are nearly everything except a kernel.
1675 This is equivalent to @option{-fno-freestanding}.
1677 @item -ffreestanding
1678 @opindex ffreestanding
1679 @cindex hosted environment
1681 Assert that compilation takes place in a freestanding environment. This
1682 implies @option{-fno-builtin}. A freestanding environment
1683 is one in which the standard library may not exist, and program startup may
1684 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1685 This is equivalent to @option{-fno-hosted}.
1687 @xref{Standards,,Language Standards Supported by GCC}, for details of
1688 freestanding and hosted environments.
1692 @cindex OpenMP parallel
1693 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1694 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1695 compiler generates parallel code according to the OpenMP Application
1696 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1697 implies @option{-pthread}, and thus is only supported on targets that
1698 have support for @option{-pthread}.
1700 @item -fms-extensions
1701 @opindex fms-extensions
1702 Accept some non-standard constructs used in Microsoft header files.
1704 In C++ code, this allows member names in structures to be similar
1705 to previous types declarations.
1714 Some cases of unnamed fields in structures and unions are only
1715 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1716 fields within structs/unions}, for details.
1718 @item -fplan9-extensions
1719 Accept some non-standard constructs used in Plan 9 code.
1721 This enables @option{-fms-extensions}, permits passing pointers to
1722 structures with anonymous fields to functions which expect pointers to
1723 elements of the type of the field, and permits referring to anonymous
1724 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1725 struct/union fields within structs/unions}, for details. This is only
1726 supported for C, not C++.
1730 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1731 options for strict ISO C conformance) implies @option{-trigraphs}.
1733 @item -no-integrated-cpp
1734 @opindex no-integrated-cpp
1735 Performs a compilation in two passes: preprocessing and compiling. This
1736 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1737 @option{-B} option. The user supplied compilation step can then add in
1738 an additional preprocessing step after normal preprocessing but before
1739 compiling. The default is to use the integrated cpp (internal cpp)
1741 The semantics of this option will change if "cc1", "cc1plus", and
1742 "cc1obj" are merged.
1744 @cindex traditional C language
1745 @cindex C language, traditional
1747 @itemx -traditional-cpp
1748 @opindex traditional-cpp
1749 @opindex traditional
1750 Formerly, these options caused GCC to attempt to emulate a pre-standard
1751 C compiler. They are now only supported with the @option{-E} switch.
1752 The preprocessor continues to support a pre-standard mode. See the GNU
1753 CPP manual for details.
1755 @item -fcond-mismatch
1756 @opindex fcond-mismatch
1757 Allow conditional expressions with mismatched types in the second and
1758 third arguments. The value of such an expression is void. This option
1759 is not supported for C++.
1761 @item -flax-vector-conversions
1762 @opindex flax-vector-conversions
1763 Allow implicit conversions between vectors with differing numbers of
1764 elements and/or incompatible element types. This option should not be
1767 @item -funsigned-char
1768 @opindex funsigned-char
1769 Let the type @code{char} be unsigned, like @code{unsigned char}.
1771 Each kind of machine has a default for what @code{char} should
1772 be. It is either like @code{unsigned char} by default or like
1773 @code{signed char} by default.
1775 Ideally, a portable program should always use @code{signed char} or
1776 @code{unsigned char} when it depends on the signedness of an object.
1777 But many programs have been written to use plain @code{char} and
1778 expect it to be signed, or expect it to be unsigned, depending on the
1779 machines they were written for. This option, and its inverse, let you
1780 make such a program work with the opposite default.
1782 The type @code{char} is always a distinct type from each of
1783 @code{signed char} or @code{unsigned char}, even though its behavior
1784 is always just like one of those two.
1787 @opindex fsigned-char
1788 Let the type @code{char} be signed, like @code{signed char}.
1790 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1791 the negative form of @option{-funsigned-char}. Likewise, the option
1792 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1794 @item -fsigned-bitfields
1795 @itemx -funsigned-bitfields
1796 @itemx -fno-signed-bitfields
1797 @itemx -fno-unsigned-bitfields
1798 @opindex fsigned-bitfields
1799 @opindex funsigned-bitfields
1800 @opindex fno-signed-bitfields
1801 @opindex fno-unsigned-bitfields
1802 These options control whether a bit-field is signed or unsigned, when the
1803 declaration does not use either @code{signed} or @code{unsigned}. By
1804 default, such a bit-field is signed, because this is consistent: the
1805 basic integer types such as @code{int} are signed types.
1808 @node C++ Dialect Options
1809 @section Options Controlling C++ Dialect
1811 @cindex compiler options, C++
1812 @cindex C++ options, command line
1813 @cindex options, C++
1814 This section describes the command-line options that are only meaningful
1815 for C++ programs; but you can also use most of the GNU compiler options
1816 regardless of what language your program is in. For example, you
1817 might compile a file @code{firstClass.C} like this:
1820 g++ -g -frepo -O -c firstClass.C
1824 In this example, only @option{-frepo} is an option meant
1825 only for C++ programs; you can use the other options with any
1826 language supported by GCC@.
1828 Here is a list of options that are @emph{only} for compiling C++ programs:
1832 @item -fabi-version=@var{n}
1833 @opindex fabi-version
1834 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1835 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1836 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1837 the version that conforms most closely to the C++ ABI specification.
1838 Therefore, the ABI obtained using version 0 will change as ABI bugs
1841 The default is version 2.
1843 Version 3 corrects an error in mangling a constant address as a
1846 Version 4 implements a standard mangling for vector types.
1848 Version 5 corrects the mangling of attribute const/volatile on
1849 function pointer types, decltype of a plain decl, and use of a
1850 function parameter in the declaration of another parameter.
1852 See also @option{-Wabi}.
1854 @item -fno-access-control
1855 @opindex fno-access-control
1856 Turn off all access checking. This switch is mainly useful for working
1857 around bugs in the access control code.
1861 Check that the pointer returned by @code{operator new} is non-null
1862 before attempting to modify the storage allocated. This check is
1863 normally unnecessary because the C++ standard specifies that
1864 @code{operator new} will only return @code{0} if it is declared
1865 @samp{throw()}, in which case the compiler will always check the
1866 return value even without this option. In all other cases, when
1867 @code{operator new} has a non-empty exception specification, memory
1868 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1869 @samp{new (nothrow)}.
1871 @item -fconserve-space
1872 @opindex fconserve-space
1873 Put uninitialized or runtime-initialized global variables into the
1874 common segment, as C does. This saves space in the executable at the
1875 cost of not diagnosing duplicate definitions. If you compile with this
1876 flag and your program mysteriously crashes after @code{main()} has
1877 completed, you may have an object that is being destroyed twice because
1878 two definitions were merged.
1880 This option is no longer useful on most targets, now that support has
1881 been added for putting variables into BSS without making them common.
1883 @item -fconstexpr-depth=@var{n}
1884 @opindex fconstexpr-depth
1885 Set the maximum nested evaluation depth for C++0x constexpr functions
1886 to @var{n}. A limit is needed to detect endless recursion during
1887 constant expression evaluation. The minimum specified by the standard
1888 is 512; G++ defaults to 1024.
1890 @item -fno-deduce-init-list
1891 @opindex fno-deduce-init-list
1892 Disable deduction of a template type parameter as
1893 std::initializer_list from a brace-enclosed initializer list, i.e.
1896 template <class T> auto forward(T t) -> decltype (realfn (t))
1903 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1907 This option is present because this deduction is an extension to the
1908 current specification in the C++0x working draft, and there was
1909 some concern about potential overload resolution problems.
1911 @item -ffriend-injection
1912 @opindex ffriend-injection
1913 Inject friend functions into the enclosing namespace, so that they are
1914 visible outside the scope of the class in which they are declared.
1915 Friend functions were documented to work this way in the old Annotated
1916 C++ Reference Manual, and versions of G++ before 4.1 always worked
1917 that way. However, in ISO C++ a friend function which is not declared
1918 in an enclosing scope can only be found using argument dependent
1919 lookup. This option causes friends to be injected as they were in
1922 This option is for compatibility, and may be removed in a future
1925 @item -fno-elide-constructors
1926 @opindex fno-elide-constructors
1927 The C++ standard allows an implementation to omit creating a temporary
1928 which is only used to initialize another object of the same type.
1929 Specifying this option disables that optimization, and forces G++ to
1930 call the copy constructor in all cases.
1932 @item -fno-enforce-eh-specs
1933 @opindex fno-enforce-eh-specs
1934 Don't generate code to check for violation of exception specifications
1935 at runtime. This option violates the C++ standard, but may be useful
1936 for reducing code size in production builds, much like defining
1937 @samp{NDEBUG}. This does not give user code permission to throw
1938 exceptions in violation of the exception specifications; the compiler
1939 will still optimize based on the specifications, so throwing an
1940 unexpected exception will result in undefined behavior.
1943 @itemx -fno-for-scope
1945 @opindex fno-for-scope
1946 If @option{-ffor-scope} is specified, the scope of variables declared in
1947 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1948 as specified by the C++ standard.
1949 If @option{-fno-for-scope} is specified, the scope of variables declared in
1950 a @i{for-init-statement} extends to the end of the enclosing scope,
1951 as was the case in old versions of G++, and other (traditional)
1952 implementations of C++.
1954 The default if neither flag is given to follow the standard,
1955 but to allow and give a warning for old-style code that would
1956 otherwise be invalid, or have different behavior.
1958 @item -fno-gnu-keywords
1959 @opindex fno-gnu-keywords
1960 Do not recognize @code{typeof} as a keyword, so that code can use this
1961 word as an identifier. You can use the keyword @code{__typeof__} instead.
1962 @option{-ansi} implies @option{-fno-gnu-keywords}.
1964 @item -fno-implicit-templates
1965 @opindex fno-implicit-templates
1966 Never emit code for non-inline templates which are instantiated
1967 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1968 @xref{Template Instantiation}, for more information.
1970 @item -fno-implicit-inline-templates
1971 @opindex fno-implicit-inline-templates
1972 Don't emit code for implicit instantiations of inline templates, either.
1973 The default is to handle inlines differently so that compiles with and
1974 without optimization will need the same set of explicit instantiations.
1976 @item -fno-implement-inlines
1977 @opindex fno-implement-inlines
1978 To save space, do not emit out-of-line copies of inline functions
1979 controlled by @samp{#pragma implementation}. This will cause linker
1980 errors if these functions are not inlined everywhere they are called.
1982 @item -fms-extensions
1983 @opindex fms-extensions
1984 Disable pedantic warnings about constructs used in MFC, such as implicit
1985 int and getting a pointer to member function via non-standard syntax.
1987 @item -fno-nonansi-builtins
1988 @opindex fno-nonansi-builtins
1989 Disable built-in declarations of functions that are not mandated by
1990 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1991 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1994 @opindex fnothrow-opt
1995 Treat a @code{throw()} exception specification as though it were a
1996 @code{noexcept} specification to reduce or eliminate the text size
1997 overhead relative to a function with no exception specification. If
1998 the function has local variables of types with non-trivial
1999 destructors, the exception specification will actually make the
2000 function smaller because the EH cleanups for those variables can be
2001 optimized away. The semantic effect is that an exception thrown out of
2002 a function with such an exception specification will result in a call
2003 to @code{terminate} rather than @code{unexpected}.
2005 @item -fno-operator-names
2006 @opindex fno-operator-names
2007 Do not treat the operator name keywords @code{and}, @code{bitand},
2008 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2009 synonyms as keywords.
2011 @item -fno-optional-diags
2012 @opindex fno-optional-diags
2013 Disable diagnostics that the standard says a compiler does not need to
2014 issue. Currently, the only such diagnostic issued by G++ is the one for
2015 a name having multiple meanings within a class.
2018 @opindex fpermissive
2019 Downgrade some diagnostics about nonconformant code from errors to
2020 warnings. Thus, using @option{-fpermissive} will allow some
2021 nonconforming code to compile.
2023 @item -fno-pretty-templates
2024 @opindex fno-pretty-templates
2025 When an error message refers to a specialization of a function
2026 template, the compiler will normally print the signature of the
2027 template followed by the template arguments and any typedefs or
2028 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2029 rather than @code{void f(int)}) so that it's clear which template is
2030 involved. When an error message refers to a specialization of a class
2031 template, the compiler will omit any template arguments which match
2032 the default template arguments for that template. If either of these
2033 behaviors make it harder to understand the error message rather than
2034 easier, using @option{-fno-pretty-templates} will disable them.
2038 Enable automatic template instantiation at link time. This option also
2039 implies @option{-fno-implicit-templates}. @xref{Template
2040 Instantiation}, for more information.
2044 Disable generation of information about every class with virtual
2045 functions for use by the C++ runtime type identification features
2046 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2047 of the language, you can save some space by using this flag. Note that
2048 exception handling uses the same information, but it will generate it as
2049 needed. The @samp{dynamic_cast} operator can still be used for casts that
2050 do not require runtime type information, i.e.@: casts to @code{void *} or to
2051 unambiguous base classes.
2055 Emit statistics about front-end processing at the end of the compilation.
2056 This information is generally only useful to the G++ development team.
2058 @item -fstrict-enums
2059 @opindex fstrict-enums
2060 Allow the compiler to optimize using the assumption that a value of
2061 enumeration type can only be one of the values of the enumeration (as
2062 defined in the C++ standard; basically, a value which can be
2063 represented in the minimum number of bits needed to represent all the
2064 enumerators). This assumption may not be valid if the program uses a
2065 cast to convert an arbitrary integer value to the enumeration type.
2067 @item -ftemplate-depth=@var{n}
2068 @opindex ftemplate-depth
2069 Set the maximum instantiation depth for template classes to @var{n}.
2070 A limit on the template instantiation depth is needed to detect
2071 endless recursions during template class instantiation. ANSI/ISO C++
2072 conforming programs must not rely on a maximum depth greater than 17
2073 (changed to 1024 in C++0x).
2075 @item -fno-threadsafe-statics
2076 @opindex fno-threadsafe-statics
2077 Do not emit the extra code to use the routines specified in the C++
2078 ABI for thread-safe initialization of local statics. You can use this
2079 option to reduce code size slightly in code that doesn't need to be
2082 @item -fuse-cxa-atexit
2083 @opindex fuse-cxa-atexit
2084 Register destructors for objects with static storage duration with the
2085 @code{__cxa_atexit} function rather than the @code{atexit} function.
2086 This option is required for fully standards-compliant handling of static
2087 destructors, but will only work if your C library supports
2088 @code{__cxa_atexit}.
2090 @item -fno-use-cxa-get-exception-ptr
2091 @opindex fno-use-cxa-get-exception-ptr
2092 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2093 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2094 if the runtime routine is not available.
2096 @item -fvisibility-inlines-hidden
2097 @opindex fvisibility-inlines-hidden
2098 This switch declares that the user does not attempt to compare
2099 pointers to inline methods where the addresses of the two functions
2100 were taken in different shared objects.
2102 The effect of this is that GCC may, effectively, mark inline methods with
2103 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2104 appear in the export table of a DSO and do not require a PLT indirection
2105 when used within the DSO@. Enabling this option can have a dramatic effect
2106 on load and link times of a DSO as it massively reduces the size of the
2107 dynamic export table when the library makes heavy use of templates.
2109 The behavior of this switch is not quite the same as marking the
2110 methods as hidden directly, because it does not affect static variables
2111 local to the function or cause the compiler to deduce that
2112 the function is defined in only one shared object.
2114 You may mark a method as having a visibility explicitly to negate the
2115 effect of the switch for that method. For example, if you do want to
2116 compare pointers to a particular inline method, you might mark it as
2117 having default visibility. Marking the enclosing class with explicit
2118 visibility will have no effect.
2120 Explicitly instantiated inline methods are unaffected by this option
2121 as their linkage might otherwise cross a shared library boundary.
2122 @xref{Template Instantiation}.
2124 @item -fvisibility-ms-compat
2125 @opindex fvisibility-ms-compat
2126 This flag attempts to use visibility settings to make GCC's C++
2127 linkage model compatible with that of Microsoft Visual Studio.
2129 The flag makes these changes to GCC's linkage model:
2133 It sets the default visibility to @code{hidden}, like
2134 @option{-fvisibility=hidden}.
2137 Types, but not their members, are not hidden by default.
2140 The One Definition Rule is relaxed for types without explicit
2141 visibility specifications which are defined in more than one different
2142 shared object: those declarations are permitted if they would have
2143 been permitted when this option was not used.
2146 In new code it is better to use @option{-fvisibility=hidden} and
2147 export those classes which are intended to be externally visible.
2148 Unfortunately it is possible for code to rely, perhaps accidentally,
2149 on the Visual Studio behavior.
2151 Among the consequences of these changes are that static data members
2152 of the same type with the same name but defined in different shared
2153 objects will be different, so changing one will not change the other;
2154 and that pointers to function members defined in different shared
2155 objects may not compare equal. When this flag is given, it is a
2156 violation of the ODR to define types with the same name differently.
2160 Do not use weak symbol support, even if it is provided by the linker.
2161 By default, G++ will use weak symbols if they are available. This
2162 option exists only for testing, and should not be used by end-users;
2163 it will result in inferior code and has no benefits. This option may
2164 be removed in a future release of G++.
2168 Do not search for header files in the standard directories specific to
2169 C++, but do still search the other standard directories. (This option
2170 is used when building the C++ library.)
2173 In addition, these optimization, warning, and code generation options
2174 have meanings only for C++ programs:
2177 @item -fno-default-inline
2178 @opindex fno-default-inline
2179 Do not assume @samp{inline} for functions defined inside a class scope.
2180 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2181 functions will have linkage like inline functions; they just won't be
2184 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2187 Warn when G++ generates code that is probably not compatible with the
2188 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2189 all such cases, there are probably some cases that are not warned about,
2190 even though G++ is generating incompatible code. There may also be
2191 cases where warnings are emitted even though the code that is generated
2194 You should rewrite your code to avoid these warnings if you are
2195 concerned about the fact that code generated by G++ may not be binary
2196 compatible with code generated by other compilers.
2198 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2203 A template with a non-type template parameter of reference type is
2204 mangled incorrectly:
2207 template <int &> struct S @{@};
2211 This is fixed in @option{-fabi-version=3}.
2214 SIMD vector types declared using @code{__attribute ((vector_size))} are
2215 mangled in a non-standard way that does not allow for overloading of
2216 functions taking vectors of different sizes.
2218 The mangling is changed in @option{-fabi-version=4}.
2221 The known incompatibilities in @option{-fabi-version=1} include:
2226 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2227 pack data into the same byte as a base class. For example:
2230 struct A @{ virtual void f(); int f1 : 1; @};
2231 struct B : public A @{ int f2 : 1; @};
2235 In this case, G++ will place @code{B::f2} into the same byte
2236 as@code{A::f1}; other compilers will not. You can avoid this problem
2237 by explicitly padding @code{A} so that its size is a multiple of the
2238 byte size on your platform; that will cause G++ and other compilers to
2239 layout @code{B} identically.
2242 Incorrect handling of tail-padding for virtual bases. G++ does not use
2243 tail padding when laying out virtual bases. For example:
2246 struct A @{ virtual void f(); char c1; @};
2247 struct B @{ B(); char c2; @};
2248 struct C : public A, public virtual B @{@};
2252 In this case, G++ will not place @code{B} into the tail-padding for
2253 @code{A}; other compilers will. You can avoid this problem by
2254 explicitly padding @code{A} so that its size is a multiple of its
2255 alignment (ignoring virtual base classes); that will cause G++ and other
2256 compilers to layout @code{C} identically.
2259 Incorrect handling of bit-fields with declared widths greater than that
2260 of their underlying types, when the bit-fields appear in a union. For
2264 union U @{ int i : 4096; @};
2268 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2269 union too small by the number of bits in an @code{int}.
2272 Empty classes can be placed at incorrect offsets. For example:
2282 struct C : public B, public A @{@};
2286 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2287 it should be placed at offset zero. G++ mistakenly believes that the
2288 @code{A} data member of @code{B} is already at offset zero.
2291 Names of template functions whose types involve @code{typename} or
2292 template template parameters can be mangled incorrectly.
2295 template <typename Q>
2296 void f(typename Q::X) @{@}
2298 template <template <typename> class Q>
2299 void f(typename Q<int>::X) @{@}
2303 Instantiations of these templates may be mangled incorrectly.
2307 It also warns psABI related changes. The known psABI changes at this
2313 For SYSV/x86-64, when passing union with long double, it is changed to
2314 pass in memory as specified in psABI. For example:
2324 @code{union U} will always be passed in memory.
2328 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2329 @opindex Wctor-dtor-privacy
2330 @opindex Wno-ctor-dtor-privacy
2331 Warn when a class seems unusable because all the constructors or
2332 destructors in that class are private, and it has neither friends nor
2333 public static member functions.
2335 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2337 @opindex Wno-noexcept
2338 Warn when a noexcept-expression evaluates to false because of a call
2339 to a function that does not have a non-throwing exception
2340 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2341 the compiler to never throw an exception.
2343 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2344 @opindex Wnon-virtual-dtor
2345 @opindex Wno-non-virtual-dtor
2346 Warn when a class has virtual functions and accessible non-virtual
2347 destructor, in which case it would be possible but unsafe to delete
2348 an instance of a derived class through a pointer to the base class.
2349 This warning is also enabled if -Weffc++ is specified.
2351 @item -Wreorder @r{(C++ and Objective-C++ only)}
2353 @opindex Wno-reorder
2354 @cindex reordering, warning
2355 @cindex warning for reordering of member initializers
2356 Warn when the order of member initializers given in the code does not
2357 match the order in which they must be executed. For instance:
2363 A(): j (0), i (1) @{ @}
2367 The compiler will rearrange the member initializers for @samp{i}
2368 and @samp{j} to match the declaration order of the members, emitting
2369 a warning to that effect. This warning is enabled by @option{-Wall}.
2372 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2375 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2378 Warn about violations of the following style guidelines from Scott Meyers'
2379 @cite{Effective C++} book:
2383 Item 11: Define a copy constructor and an assignment operator for classes
2384 with dynamically allocated memory.
2387 Item 12: Prefer initialization to assignment in constructors.
2390 Item 14: Make destructors virtual in base classes.
2393 Item 15: Have @code{operator=} return a reference to @code{*this}.
2396 Item 23: Don't try to return a reference when you must return an object.
2400 Also warn about violations of the following style guidelines from
2401 Scott Meyers' @cite{More Effective C++} book:
2405 Item 6: Distinguish between prefix and postfix forms of increment and
2406 decrement operators.
2409 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2413 When selecting this option, be aware that the standard library
2414 headers do not obey all of these guidelines; use @samp{grep -v}
2415 to filter out those warnings.
2417 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2418 @opindex Wstrict-null-sentinel
2419 @opindex Wno-strict-null-sentinel
2420 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2421 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2422 to @code{__null}. Although it is a null pointer constant not a null pointer,
2423 it is guaranteed to be of the same size as a pointer. But this use is
2424 not portable across different compilers.
2426 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2427 @opindex Wno-non-template-friend
2428 @opindex Wnon-template-friend
2429 Disable warnings when non-templatized friend functions are declared
2430 within a template. Since the advent of explicit template specification
2431 support in G++, if the name of the friend is an unqualified-id (i.e.,
2432 @samp{friend foo(int)}), the C++ language specification demands that the
2433 friend declare or define an ordinary, nontemplate function. (Section
2434 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2435 could be interpreted as a particular specialization of a templatized
2436 function. Because this non-conforming behavior is no longer the default
2437 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2438 check existing code for potential trouble spots and is on by default.
2439 This new compiler behavior can be turned off with
2440 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2441 but disables the helpful warning.
2443 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2444 @opindex Wold-style-cast
2445 @opindex Wno-old-style-cast
2446 Warn if an old-style (C-style) cast to a non-void type is used within
2447 a C++ program. The new-style casts (@samp{dynamic_cast},
2448 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2449 less vulnerable to unintended effects and much easier to search for.
2451 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2452 @opindex Woverloaded-virtual
2453 @opindex Wno-overloaded-virtual
2454 @cindex overloaded virtual function, warning
2455 @cindex warning for overloaded virtual function
2456 Warn when a function declaration hides virtual functions from a
2457 base class. For example, in:
2464 struct B: public A @{
2469 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2477 will fail to compile.
2479 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2480 @opindex Wno-pmf-conversions
2481 @opindex Wpmf-conversions
2482 Disable the diagnostic for converting a bound pointer to member function
2485 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2486 @opindex Wsign-promo
2487 @opindex Wno-sign-promo
2488 Warn when overload resolution chooses a promotion from unsigned or
2489 enumerated type to a signed type, over a conversion to an unsigned type of
2490 the same size. Previous versions of G++ would try to preserve
2491 unsignedness, but the standard mandates the current behavior.
2496 A& operator = (int);
2506 In this example, G++ will synthesize a default @samp{A& operator =
2507 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2510 @node Objective-C and Objective-C++ Dialect Options
2511 @section Options Controlling Objective-C and Objective-C++ Dialects
2513 @cindex compiler options, Objective-C and Objective-C++
2514 @cindex Objective-C and Objective-C++ options, command line
2515 @cindex options, Objective-C and Objective-C++
2516 (NOTE: This manual does not describe the Objective-C and Objective-C++
2517 languages themselves. @xref{Standards,,Language Standards
2518 Supported by GCC}, for references.)
2520 This section describes the command-line options that are only meaningful
2521 for Objective-C and Objective-C++ programs, but you can also use most of
2522 the language-independent GNU compiler options.
2523 For example, you might compile a file @code{some_class.m} like this:
2526 gcc -g -fgnu-runtime -O -c some_class.m
2530 In this example, @option{-fgnu-runtime} is an option meant only for
2531 Objective-C and Objective-C++ programs; you can use the other options with
2532 any language supported by GCC@.
2534 Note that since Objective-C is an extension of the C language, Objective-C
2535 compilations may also use options specific to the C front-end (e.g.,
2536 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2537 C++-specific options (e.g., @option{-Wabi}).
2539 Here is a list of options that are @emph{only} for compiling Objective-C
2540 and Objective-C++ programs:
2543 @item -fconstant-string-class=@var{class-name}
2544 @opindex fconstant-string-class
2545 Use @var{class-name} as the name of the class to instantiate for each
2546 literal string specified with the syntax @code{@@"@dots{}"}. The default
2547 class name is @code{NXConstantString} if the GNU runtime is being used, and
2548 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2549 @option{-fconstant-cfstrings} option, if also present, will override the
2550 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2551 to be laid out as constant CoreFoundation strings.
2554 @opindex fgnu-runtime
2555 Generate object code compatible with the standard GNU Objective-C
2556 runtime. This is the default for most types of systems.
2558 @item -fnext-runtime
2559 @opindex fnext-runtime
2560 Generate output compatible with the NeXT runtime. This is the default
2561 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2562 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2565 @item -fno-nil-receivers
2566 @opindex fno-nil-receivers
2567 Assume that all Objective-C message dispatches (@code{[receiver
2568 message:arg]}) in this translation unit ensure that the receiver is
2569 not @code{nil}. This allows for more efficient entry points in the
2570 runtime to be used. This option is only available in conjunction with
2571 the NeXT runtime and ABI version 0 or 1.
2573 @item -fobjc-abi-version=@var{n}
2574 @opindex fobjc-abi-version
2575 Use version @var{n} of the Objective-C ABI for the selected runtime.
2576 This option is currently supported only for the NeXT runtime. In that
2577 case, Version 0 is the traditional (32-bit) ABI without support for
2578 properties and other Objective-C 2.0 additions. Version 1 is the
2579 traditional (32-bit) ABI with support for properties and other
2580 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2581 nothing is specified, the default is Version 0 on 32-bit target
2582 machines, and Version 2 on 64-bit target machines.
2584 @item -fobjc-call-cxx-cdtors
2585 @opindex fobjc-call-cxx-cdtors
2586 For each Objective-C class, check if any of its instance variables is a
2587 C++ object with a non-trivial default constructor. If so, synthesize a
2588 special @code{- (id) .cxx_construct} instance method that will run
2589 non-trivial default constructors on any such instance variables, in order,
2590 and then return @code{self}. Similarly, check if any instance variable
2591 is a C++ object with a non-trivial destructor, and if so, synthesize a
2592 special @code{- (void) .cxx_destruct} method that will run
2593 all such default destructors, in reverse order.
2595 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2596 methods thusly generated will only operate on instance variables
2597 declared in the current Objective-C class, and not those inherited
2598 from superclasses. It is the responsibility of the Objective-C
2599 runtime to invoke all such methods in an object's inheritance
2600 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2601 by the runtime immediately after a new object instance is allocated;
2602 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2603 before the runtime deallocates an object instance.
2605 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2606 support for invoking the @code{- (id) .cxx_construct} and
2607 @code{- (void) .cxx_destruct} methods.
2609 @item -fobjc-direct-dispatch
2610 @opindex fobjc-direct-dispatch
2611 Allow fast jumps to the message dispatcher. On Darwin this is
2612 accomplished via the comm page.
2614 @item -fobjc-exceptions
2615 @opindex fobjc-exceptions
2616 Enable syntactic support for structured exception handling in
2617 Objective-C, similar to what is offered by C++ and Java. This option
2618 is required to use the Objective-C keywords @code{@@try},
2619 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2620 @code{@@synchronized}. This option is available with both the GNU
2621 runtime and the NeXT runtime (but not available in conjunction with
2622 the NeXT runtime on Mac OS X 10.2 and earlier).
2626 Enable garbage collection (GC) in Objective-C and Objective-C++
2627 programs. This option is only available with the NeXT runtime; the
2628 GNU runtime has a different garbage collection implementation that
2629 does not require special compiler flags.
2631 @item -fobjc-nilcheck
2632 @opindex fobjc-nilcheck
2633 For the NeXT runtime with version 2 of the ABI, check for a nil
2634 receiver in method invocations before doing the actual method call.
2635 This is the default and can be disabled using
2636 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2637 checked for nil in this way no matter what this flag is set to.
2638 Currently this flag does nothing when the GNU runtime, or an older
2639 version of the NeXT runtime ABI, is used.
2641 @item -fobjc-std=objc1
2643 Conform to the language syntax of Objective-C 1.0, the language
2644 recognized by GCC 4.0. This only affects the Objective-C additions to
2645 the C/C++ language; it does not affect conformance to C/C++ standards,
2646 which is controlled by the separate C/C++ dialect option flags. When
2647 this option is used with the Objective-C or Objective-C++ compiler,
2648 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2649 This is useful if you need to make sure that your Objective-C code can
2650 be compiled with older versions of GCC.
2652 @item -freplace-objc-classes
2653 @opindex freplace-objc-classes
2654 Emit a special marker instructing @command{ld(1)} not to statically link in
2655 the resulting object file, and allow @command{dyld(1)} to load it in at
2656 run time instead. This is used in conjunction with the Fix-and-Continue
2657 debugging mode, where the object file in question may be recompiled and
2658 dynamically reloaded in the course of program execution, without the need
2659 to restart the program itself. Currently, Fix-and-Continue functionality
2660 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2665 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2666 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2667 compile time) with static class references that get initialized at load time,
2668 which improves run-time performance. Specifying the @option{-fzero-link} flag
2669 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2670 to be retained. This is useful in Zero-Link debugging mode, since it allows
2671 for individual class implementations to be modified during program execution.
2672 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2673 regardless of command line options.
2677 Dump interface declarations for all classes seen in the source file to a
2678 file named @file{@var{sourcename}.decl}.
2680 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2681 @opindex Wassign-intercept
2682 @opindex Wno-assign-intercept
2683 Warn whenever an Objective-C assignment is being intercepted by the
2686 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2687 @opindex Wno-protocol
2689 If a class is declared to implement a protocol, a warning is issued for
2690 every method in the protocol that is not implemented by the class. The
2691 default behavior is to issue a warning for every method not explicitly
2692 implemented in the class, even if a method implementation is inherited
2693 from the superclass. If you use the @option{-Wno-protocol} option, then
2694 methods inherited from the superclass are considered to be implemented,
2695 and no warning is issued for them.
2697 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2699 @opindex Wno-selector
2700 Warn if multiple methods of different types for the same selector are
2701 found during compilation. The check is performed on the list of methods
2702 in the final stage of compilation. Additionally, a check is performed
2703 for each selector appearing in a @code{@@selector(@dots{})}
2704 expression, and a corresponding method for that selector has been found
2705 during compilation. Because these checks scan the method table only at
2706 the end of compilation, these warnings are not produced if the final
2707 stage of compilation is not reached, for example because an error is
2708 found during compilation, or because the @option{-fsyntax-only} option is
2711 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2712 @opindex Wstrict-selector-match
2713 @opindex Wno-strict-selector-match
2714 Warn if multiple methods with differing argument and/or return types are
2715 found for a given selector when attempting to send a message using this
2716 selector to a receiver of type @code{id} or @code{Class}. When this flag
2717 is off (which is the default behavior), the compiler will omit such warnings
2718 if any differences found are confined to types which share the same size
2721 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2722 @opindex Wundeclared-selector
2723 @opindex Wno-undeclared-selector
2724 Warn if a @code{@@selector(@dots{})} expression referring to an
2725 undeclared selector is found. A selector is considered undeclared if no
2726 method with that name has been declared before the
2727 @code{@@selector(@dots{})} expression, either explicitly in an
2728 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2729 an @code{@@implementation} section. This option always performs its
2730 checks as soon as a @code{@@selector(@dots{})} expression is found,
2731 while @option{-Wselector} only performs its checks in the final stage of
2732 compilation. This also enforces the coding style convention
2733 that methods and selectors must be declared before being used.
2735 @item -print-objc-runtime-info
2736 @opindex print-objc-runtime-info
2737 Generate C header describing the largest structure that is passed by
2742 @node Language Independent Options
2743 @section Options to Control Diagnostic Messages Formatting
2744 @cindex options to control diagnostics formatting
2745 @cindex diagnostic messages
2746 @cindex message formatting
2748 Traditionally, diagnostic messages have been formatted irrespective of
2749 the output device's aspect (e.g.@: its width, @dots{}). The options described
2750 below can be used to control the diagnostic messages formatting
2751 algorithm, e.g.@: how many characters per line, how often source location
2752 information should be reported. Right now, only the C++ front end can
2753 honor these options. However it is expected, in the near future, that
2754 the remaining front ends would be able to digest them correctly.
2757 @item -fmessage-length=@var{n}
2758 @opindex fmessage-length
2759 Try to format error messages so that they fit on lines of about @var{n}
2760 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2761 the front ends supported by GCC@. If @var{n} is zero, then no
2762 line-wrapping will be done; each error message will appear on a single
2765 @opindex fdiagnostics-show-location
2766 @item -fdiagnostics-show-location=once
2767 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2768 reporter to emit @emph{once} source location information; that is, in
2769 case the message is too long to fit on a single physical line and has to
2770 be wrapped, the source location won't be emitted (as prefix) again,
2771 over and over, in subsequent continuation lines. This is the default
2774 @item -fdiagnostics-show-location=every-line
2775 Only meaningful in line-wrapping mode. Instructs the diagnostic
2776 messages reporter to emit the same source location information (as
2777 prefix) for physical lines that result from the process of breaking
2778 a message which is too long to fit on a single line.
2780 @item -fno-diagnostics-show-option
2781 @opindex fno-diagnostics-show-option
2782 @opindex fdiagnostics-show-option
2783 By default, each diagnostic emitted includes text which indicates the
2784 command line option that directly controls the diagnostic (if such an
2785 option is known to the diagnostic machinery). Specifying the
2786 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
2788 @item -Wcoverage-mismatch
2789 @opindex Wcoverage-mismatch
2790 Warn if feedback profiles do not match when using the
2791 @option{-fprofile-use} option.
2792 If a source file was changed between @option{-fprofile-gen} and
2793 @option{-fprofile-use}, the files with the profile feedback can fail
2794 to match the source file and GCC can not use the profile feedback
2795 information. By default, this warning is enabled and is treated as an
2796 error. @option{-Wno-coverage-mismatch} can be used to disable the
2797 warning or @option{-Wno-error=coverage-mismatch} can be used to
2798 disable the error. Disable the error for this warning can result in
2799 poorly optimized code, so disabling the error is useful only in the
2800 case of very minor changes such as bug fixes to an existing code-base.
2801 Completely disabling the warning is not recommended.
2805 @node Warning Options
2806 @section Options to Request or Suppress Warnings
2807 @cindex options to control warnings
2808 @cindex warning messages
2809 @cindex messages, warning
2810 @cindex suppressing warnings
2812 Warnings are diagnostic messages that report constructions which
2813 are not inherently erroneous but which are risky or suggest there
2814 may have been an error.
2816 The following language-independent options do not enable specific
2817 warnings but control the kinds of diagnostics produced by GCC.
2820 @cindex syntax checking
2822 @opindex fsyntax-only
2823 Check the code for syntax errors, but don't do anything beyond that.
2825 @item -fmax-errors=@var{n}
2826 @opindex fmax-errors
2827 Limits the maximum number of error messages to @var{n}, at which point
2828 GCC bails out rather than attempting to continue processing the source
2829 code. If @var{n} is 0 (the default), there is no limit on the number
2830 of error messages produced. If @option{-Wfatal-errors} is also
2831 specified, then @option{-Wfatal-errors} takes precedence over this
2836 Inhibit all warning messages.
2841 Make all warnings into errors.
2846 Make the specified warning into an error. The specifier for a warning
2847 is appended, for example @option{-Werror=switch} turns the warnings
2848 controlled by @option{-Wswitch} into errors. This switch takes a
2849 negative form, to be used to negate @option{-Werror} for specific
2850 warnings, for example @option{-Wno-error=switch} makes
2851 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2854 The warning message for each controllable warning includes the
2855 option which controls the warning. That option can then be used with
2856 @option{-Werror=} and @option{-Wno-error=} as described above.
2857 (Printing of the option in the warning message can be disabled using the
2858 @option{-fno-diagnostics-show-option} flag.)
2860 Note that specifying @option{-Werror=}@var{foo} automatically implies
2861 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2864 @item -Wfatal-errors
2865 @opindex Wfatal-errors
2866 @opindex Wno-fatal-errors
2867 This option causes the compiler to abort compilation on the first error
2868 occurred rather than trying to keep going and printing further error
2873 You can request many specific warnings with options beginning
2874 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2875 implicit declarations. Each of these specific warning options also
2876 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2877 example, @option{-Wno-implicit}. This manual lists only one of the
2878 two forms, whichever is not the default. For further,
2879 language-specific options also refer to @ref{C++ Dialect Options} and
2880 @ref{Objective-C and Objective-C++ Dialect Options}.
2882 When an unrecognized warning option is requested (e.g.,
2883 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2884 that the option is not recognized. However, if the @option{-Wno-} form
2885 is used, the behavior is slightly different: No diagnostic will be
2886 produced for @option{-Wno-unknown-warning} unless other diagnostics
2887 are being produced. This allows the use of new @option{-Wno-} options
2888 with old compilers, but if something goes wrong, the compiler will
2889 warn that an unrecognized option was used.
2894 Issue all the warnings demanded by strict ISO C and ISO C++;
2895 reject all programs that use forbidden extensions, and some other
2896 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2897 version of the ISO C standard specified by any @option{-std} option used.
2899 Valid ISO C and ISO C++ programs should compile properly with or without
2900 this option (though a rare few will require @option{-ansi} or a
2901 @option{-std} option specifying the required version of ISO C)@. However,
2902 without this option, certain GNU extensions and traditional C and C++
2903 features are supported as well. With this option, they are rejected.
2905 @option{-pedantic} does not cause warning messages for use of the
2906 alternate keywords whose names begin and end with @samp{__}. Pedantic
2907 warnings are also disabled in the expression that follows
2908 @code{__extension__}. However, only system header files should use
2909 these escape routes; application programs should avoid them.
2910 @xref{Alternate Keywords}.
2912 Some users try to use @option{-pedantic} to check programs for strict ISO
2913 C conformance. They soon find that it does not do quite what they want:
2914 it finds some non-ISO practices, but not all---only those for which
2915 ISO C @emph{requires} a diagnostic, and some others for which
2916 diagnostics have been added.
2918 A feature to report any failure to conform to ISO C might be useful in
2919 some instances, but would require considerable additional work and would
2920 be quite different from @option{-pedantic}. We don't have plans to
2921 support such a feature in the near future.
2923 Where the standard specified with @option{-std} represents a GNU
2924 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2925 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2926 extended dialect is based. Warnings from @option{-pedantic} are given
2927 where they are required by the base standard. (It would not make sense
2928 for such warnings to be given only for features not in the specified GNU
2929 C dialect, since by definition the GNU dialects of C include all
2930 features the compiler supports with the given option, and there would be
2931 nothing to warn about.)
2933 @item -pedantic-errors
2934 @opindex pedantic-errors
2935 Like @option{-pedantic}, except that errors are produced rather than
2941 This enables all the warnings about constructions that some users
2942 consider questionable, and that are easy to avoid (or modify to
2943 prevent the warning), even in conjunction with macros. This also
2944 enables some language-specific warnings described in @ref{C++ Dialect
2945 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2947 @option{-Wall} turns on the following warning flags:
2949 @gccoptlist{-Waddress @gol
2950 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2952 -Wchar-subscripts @gol
2953 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2954 -Wimplicit-int @r{(C and Objective-C only)} @gol
2955 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
2958 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2959 -Wmissing-braces @gol
2965 -Wsequence-point @gol
2966 -Wsign-compare @r{(only in C++)} @gol
2967 -Wstrict-aliasing @gol
2968 -Wstrict-overflow=1 @gol
2971 -Wuninitialized @gol
2972 -Wunknown-pragmas @gol
2973 -Wunused-function @gol
2976 -Wunused-variable @gol
2977 -Wvolatile-register-var @gol
2980 Note that some warning flags are not implied by @option{-Wall}. Some of
2981 them warn about constructions that users generally do not consider
2982 questionable, but which occasionally you might wish to check for;
2983 others warn about constructions that are necessary or hard to avoid in
2984 some cases, and there is no simple way to modify the code to suppress
2985 the warning. Some of them are enabled by @option{-Wextra} but many of
2986 them must be enabled individually.
2992 This enables some extra warning flags that are not enabled by
2993 @option{-Wall}. (This option used to be called @option{-W}. The older
2994 name is still supported, but the newer name is more descriptive.)
2996 @gccoptlist{-Wclobbered @gol
2998 -Wignored-qualifiers @gol
2999 -Wmissing-field-initializers @gol
3000 -Wmissing-parameter-type @r{(C only)} @gol
3001 -Wold-style-declaration @r{(C only)} @gol
3002 -Woverride-init @gol
3005 -Wuninitialized @gol
3006 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3007 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3010 The option @option{-Wextra} also prints warning messages for the
3016 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3017 @samp{>}, or @samp{>=}.
3020 (C++ only) An enumerator and a non-enumerator both appear in a
3021 conditional expression.
3024 (C++ only) Ambiguous virtual bases.
3027 (C++ only) Subscripting an array which has been declared @samp{register}.
3030 (C++ only) Taking the address of a variable which has been declared
3034 (C++ only) A base class is not initialized in a derived class' copy
3039 @item -Wchar-subscripts
3040 @opindex Wchar-subscripts
3041 @opindex Wno-char-subscripts
3042 Warn if an array subscript has type @code{char}. This is a common cause
3043 of error, as programmers often forget that this type is signed on some
3045 This warning is enabled by @option{-Wall}.
3049 @opindex Wno-comment
3050 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3051 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3052 This warning is enabled by @option{-Wall}.
3055 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3057 Suppress warning messages emitted by @code{#warning} directives.
3059 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3060 @opindex Wdouble-promotion
3061 @opindex Wno-double-promotion
3062 Give a warning when a value of type @code{float} is implicitly
3063 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3064 floating-point unit implement @code{float} in hardware, but emulate
3065 @code{double} in software. On such a machine, doing computations
3066 using @code{double} values is much more expensive because of the
3067 overhead required for software emulation.
3069 It is easy to accidentally do computations with @code{double} because
3070 floating-point literals are implicitly of type @code{double}. For
3074 float area(float radius)
3076 return 3.14159 * radius * radius;
3080 the compiler will perform the entire computation with @code{double}
3081 because the floating-point literal is a @code{double}.
3086 @opindex ffreestanding
3087 @opindex fno-builtin
3088 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3089 the arguments supplied have types appropriate to the format string
3090 specified, and that the conversions specified in the format string make
3091 sense. This includes standard functions, and others specified by format
3092 attributes (@pxref{Function Attributes}), in the @code{printf},
3093 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3094 not in the C standard) families (or other target-specific families).
3095 Which functions are checked without format attributes having been
3096 specified depends on the standard version selected, and such checks of
3097 functions without the attribute specified are disabled by
3098 @option{-ffreestanding} or @option{-fno-builtin}.
3100 The formats are checked against the format features supported by GNU
3101 libc version 2.2. These include all ISO C90 and C99 features, as well
3102 as features from the Single Unix Specification and some BSD and GNU
3103 extensions. Other library implementations may not support all these
3104 features; GCC does not support warning about features that go beyond a
3105 particular library's limitations. However, if @option{-pedantic} is used
3106 with @option{-Wformat}, warnings will be given about format features not
3107 in the selected standard version (but not for @code{strfmon} formats,
3108 since those are not in any version of the C standard). @xref{C Dialect
3109 Options,,Options Controlling C Dialect}.
3111 Since @option{-Wformat} also checks for null format arguments for
3112 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3114 @option{-Wformat} is included in @option{-Wall}. For more control over some
3115 aspects of format checking, the options @option{-Wformat-y2k},
3116 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3117 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3118 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3121 @opindex Wformat-y2k
3122 @opindex Wno-format-y2k
3123 If @option{-Wformat} is specified, also warn about @code{strftime}
3124 formats which may yield only a two-digit year.
3126 @item -Wno-format-contains-nul
3127 @opindex Wno-format-contains-nul
3128 @opindex Wformat-contains-nul
3129 If @option{-Wformat} is specified, do not warn about format strings that
3132 @item -Wno-format-extra-args
3133 @opindex Wno-format-extra-args
3134 @opindex Wformat-extra-args
3135 If @option{-Wformat} is specified, do not warn about excess arguments to a
3136 @code{printf} or @code{scanf} format function. The C standard specifies
3137 that such arguments are ignored.
3139 Where the unused arguments lie between used arguments that are
3140 specified with @samp{$} operand number specifications, normally
3141 warnings are still given, since the implementation could not know what
3142 type to pass to @code{va_arg} to skip the unused arguments. However,
3143 in the case of @code{scanf} formats, this option will suppress the
3144 warning if the unused arguments are all pointers, since the Single
3145 Unix Specification says that such unused arguments are allowed.
3147 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3148 @opindex Wno-format-zero-length
3149 @opindex Wformat-zero-length
3150 If @option{-Wformat} is specified, do not warn about zero-length formats.
3151 The C standard specifies that zero-length formats are allowed.
3153 @item -Wformat-nonliteral
3154 @opindex Wformat-nonliteral
3155 @opindex Wno-format-nonliteral
3156 If @option{-Wformat} is specified, also warn if the format string is not a
3157 string literal and so cannot be checked, unless the format function
3158 takes its format arguments as a @code{va_list}.
3160 @item -Wformat-security
3161 @opindex Wformat-security
3162 @opindex Wno-format-security
3163 If @option{-Wformat} is specified, also warn about uses of format
3164 functions that represent possible security problems. At present, this
3165 warns about calls to @code{printf} and @code{scanf} functions where the
3166 format string is not a string literal and there are no format arguments,
3167 as in @code{printf (foo);}. This may be a security hole if the format
3168 string came from untrusted input and contains @samp{%n}. (This is
3169 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3170 in future warnings may be added to @option{-Wformat-security} that are not
3171 included in @option{-Wformat-nonliteral}.)
3175 @opindex Wno-format=2
3176 Enable @option{-Wformat} plus format checks not included in
3177 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3178 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3180 @item -Wnonnull @r{(C and Objective-C only)}
3182 @opindex Wno-nonnull
3183 Warn about passing a null pointer for arguments marked as
3184 requiring a non-null value by the @code{nonnull} function attribute.
3186 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3187 can be disabled with the @option{-Wno-nonnull} option.
3189 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3191 @opindex Wno-init-self
3192 Warn about uninitialized variables which are initialized with themselves.
3193 Note this option can only be used with the @option{-Wuninitialized} option.
3195 For example, GCC will warn about @code{i} being uninitialized in the
3196 following snippet only when @option{-Winit-self} has been specified:
3207 @item -Wimplicit-int @r{(C and Objective-C only)}
3208 @opindex Wimplicit-int
3209 @opindex Wno-implicit-int
3210 Warn when a declaration does not specify a type.
3211 This warning is enabled by @option{-Wall}.
3213 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3214 @opindex Wimplicit-function-declaration
3215 @opindex Wno-implicit-function-declaration
3216 Give a warning whenever a function is used before being declared. In
3217 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3218 enabled by default and it is made into an error by
3219 @option{-pedantic-errors}. This warning is also enabled by
3222 @item -Wimplicit @r{(C and Objective-C only)}
3224 @opindex Wno-implicit
3225 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3226 This warning is enabled by @option{-Wall}.
3228 @item -Wignored-qualifiers @r{(C and C++ only)}
3229 @opindex Wignored-qualifiers
3230 @opindex Wno-ignored-qualifiers
3231 Warn if the return type of a function has a type qualifier
3232 such as @code{const}. For ISO C such a type qualifier has no effect,
3233 since the value returned by a function is not an lvalue.
3234 For C++, the warning is only emitted for scalar types or @code{void}.
3235 ISO C prohibits qualified @code{void} return types on function
3236 definitions, so such return types always receive a warning
3237 even without this option.
3239 This warning is also enabled by @option{-Wextra}.
3244 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3245 a function with external linkage, returning int, taking either zero
3246 arguments, two, or three arguments of appropriate types. This warning
3247 is enabled by default in C++ and is enabled by either @option{-Wall}
3248 or @option{-pedantic}.
3250 @item -Wmissing-braces
3251 @opindex Wmissing-braces
3252 @opindex Wno-missing-braces
3253 Warn if an aggregate or union initializer is not fully bracketed. In
3254 the following example, the initializer for @samp{a} is not fully
3255 bracketed, but that for @samp{b} is fully bracketed.
3258 int a[2][2] = @{ 0, 1, 2, 3 @};
3259 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3262 This warning is enabled by @option{-Wall}.
3264 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3265 @opindex Wmissing-include-dirs
3266 @opindex Wno-missing-include-dirs
3267 Warn if a user-supplied include directory does not exist.
3270 @opindex Wparentheses
3271 @opindex Wno-parentheses
3272 Warn if parentheses are omitted in certain contexts, such
3273 as when there is an assignment in a context where a truth value
3274 is expected, or when operators are nested whose precedence people
3275 often get confused about.
3277 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3278 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3279 interpretation from that of ordinary mathematical notation.
3281 Also warn about constructions where there may be confusion to which
3282 @code{if} statement an @code{else} branch belongs. Here is an example of
3297 In C/C++, every @code{else} branch belongs to the innermost possible
3298 @code{if} statement, which in this example is @code{if (b)}. This is
3299 often not what the programmer expected, as illustrated in the above
3300 example by indentation the programmer chose. When there is the
3301 potential for this confusion, GCC will issue a warning when this flag
3302 is specified. To eliminate the warning, add explicit braces around
3303 the innermost @code{if} statement so there is no way the @code{else}
3304 could belong to the enclosing @code{if}. The resulting code would
3321 Also warn for dangerous uses of the
3322 ?: with omitted middle operand GNU extension. When the condition
3323 in the ?: operator is a boolean expression the omitted value will
3324 be always 1. Often the user expects it to be a value computed
3325 inside the conditional expression instead.
3327 This warning is enabled by @option{-Wall}.
3329 @item -Wsequence-point
3330 @opindex Wsequence-point
3331 @opindex Wno-sequence-point
3332 Warn about code that may have undefined semantics because of violations
3333 of sequence point rules in the C and C++ standards.
3335 The C and C++ standards defines the order in which expressions in a C/C++
3336 program are evaluated in terms of @dfn{sequence points}, which represent
3337 a partial ordering between the execution of parts of the program: those
3338 executed before the sequence point, and those executed after it. These
3339 occur after the evaluation of a full expression (one which is not part
3340 of a larger expression), after the evaluation of the first operand of a
3341 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3342 function is called (but after the evaluation of its arguments and the
3343 expression denoting the called function), and in certain other places.
3344 Other than as expressed by the sequence point rules, the order of
3345 evaluation of subexpressions of an expression is not specified. All
3346 these rules describe only a partial order rather than a total order,
3347 since, for example, if two functions are called within one expression
3348 with no sequence point between them, the order in which the functions
3349 are called is not specified. However, the standards committee have
3350 ruled that function calls do not overlap.
3352 It is not specified when between sequence points modifications to the
3353 values of objects take effect. Programs whose behavior depends on this
3354 have undefined behavior; the C and C++ standards specify that ``Between
3355 the previous and next sequence point an object shall have its stored
3356 value modified at most once by the evaluation of an expression.
3357 Furthermore, the prior value shall be read only to determine the value
3358 to be stored.''. If a program breaks these rules, the results on any
3359 particular implementation are entirely unpredictable.
3361 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3362 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3363 diagnosed by this option, and it may give an occasional false positive
3364 result, but in general it has been found fairly effective at detecting
3365 this sort of problem in programs.
3367 The standard is worded confusingly, therefore there is some debate
3368 over the precise meaning of the sequence point rules in subtle cases.
3369 Links to discussions of the problem, including proposed formal
3370 definitions, may be found on the GCC readings page, at
3371 @uref{http://gcc.gnu.org/@/readings.html}.
3373 This warning is enabled by @option{-Wall} for C and C++.
3376 @opindex Wreturn-type
3377 @opindex Wno-return-type
3378 Warn whenever a function is defined with a return-type that defaults
3379 to @code{int}. Also warn about any @code{return} statement with no
3380 return-value in a function whose return-type is not @code{void}
3381 (falling off the end of the function body is considered returning
3382 without a value), and about a @code{return} statement with an
3383 expression in a function whose return-type is @code{void}.
3385 For C++, a function without return type always produces a diagnostic
3386 message, even when @option{-Wno-return-type} is specified. The only
3387 exceptions are @samp{main} and functions defined in system headers.
3389 This warning is enabled by @option{-Wall}.
3394 Warn whenever a @code{switch} statement has an index of enumerated type
3395 and lacks a @code{case} for one or more of the named codes of that
3396 enumeration. (The presence of a @code{default} label prevents this
3397 warning.) @code{case} labels outside the enumeration range also
3398 provoke warnings when this option is used (even if there is a
3399 @code{default} label).
3400 This warning is enabled by @option{-Wall}.
3402 @item -Wswitch-default
3403 @opindex Wswitch-default
3404 @opindex Wno-switch-default
3405 Warn whenever a @code{switch} statement does not have a @code{default}
3409 @opindex Wswitch-enum
3410 @opindex Wno-switch-enum
3411 Warn whenever a @code{switch} statement has an index of enumerated type
3412 and lacks a @code{case} for one or more of the named codes of that
3413 enumeration. @code{case} labels outside the enumeration range also
3414 provoke warnings when this option is used. The only difference
3415 between @option{-Wswitch} and this option is that this option gives a
3416 warning about an omitted enumeration code even if there is a
3417 @code{default} label.
3419 @item -Wsync-nand @r{(C and C++ only)}
3421 @opindex Wno-sync-nand
3422 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3423 built-in functions are used. These functions changed semantics in GCC 4.4.
3427 @opindex Wno-trigraphs
3428 Warn if any trigraphs are encountered that might change the meaning of
3429 the program (trigraphs within comments are not warned about).
3430 This warning is enabled by @option{-Wall}.
3432 @item -Wunused-but-set-parameter
3433 @opindex Wunused-but-set-parameter
3434 @opindex Wno-unused-but-set-parameter
3435 Warn whenever a function parameter is assigned to, but otherwise unused
3436 (aside from its declaration).
3438 To suppress this warning use the @samp{unused} attribute
3439 (@pxref{Variable Attributes}).
3441 This warning is also enabled by @option{-Wunused} together with
3444 @item -Wunused-but-set-variable
3445 @opindex Wunused-but-set-variable
3446 @opindex Wno-unused-but-set-variable
3447 Warn whenever a local variable is assigned to, but otherwise unused
3448 (aside from its declaration).
3449 This warning is enabled by @option{-Wall}.
3451 To suppress this warning use the @samp{unused} attribute
3452 (@pxref{Variable Attributes}).
3454 This warning is also enabled by @option{-Wunused}, which is enabled
3457 @item -Wunused-function
3458 @opindex Wunused-function
3459 @opindex Wno-unused-function
3460 Warn whenever a static function is declared but not defined or a
3461 non-inline static function is unused.
3462 This warning is enabled by @option{-Wall}.
3464 @item -Wunused-label
3465 @opindex Wunused-label
3466 @opindex Wno-unused-label
3467 Warn whenever a label is declared but not used.
3468 This warning is enabled by @option{-Wall}.
3470 To suppress this warning use the @samp{unused} attribute
3471 (@pxref{Variable Attributes}).
3473 @item -Wunused-parameter
3474 @opindex Wunused-parameter
3475 @opindex Wno-unused-parameter
3476 Warn whenever a function parameter is unused aside from its declaration.
3478 To suppress this warning use the @samp{unused} attribute
3479 (@pxref{Variable Attributes}).
3481 @item -Wno-unused-result
3482 @opindex Wunused-result
3483 @opindex Wno-unused-result
3484 Do not warn if a caller of a function marked with attribute
3485 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3486 its return value. The default is @option{-Wunused-result}.
3488 @item -Wunused-variable
3489 @opindex Wunused-variable
3490 @opindex Wno-unused-variable
3491 Warn whenever a local variable or non-constant static variable is unused
3492 aside from its declaration.
3493 This warning is enabled by @option{-Wall}.
3495 To suppress this warning use the @samp{unused} attribute
3496 (@pxref{Variable Attributes}).
3498 @item -Wunused-value
3499 @opindex Wunused-value
3500 @opindex Wno-unused-value
3501 Warn whenever a statement computes a result that is explicitly not
3502 used. To suppress this warning cast the unused expression to
3503 @samp{void}. This includes an expression-statement or the left-hand
3504 side of a comma expression that contains no side effects. For example,
3505 an expression such as @samp{x[i,j]} will cause a warning, while
3506 @samp{x[(void)i,j]} will not.
3508 This warning is enabled by @option{-Wall}.
3513 All the above @option{-Wunused} options combined.
3515 In order to get a warning about an unused function parameter, you must
3516 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3517 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3519 @item -Wuninitialized
3520 @opindex Wuninitialized
3521 @opindex Wno-uninitialized
3522 Warn if an automatic variable is used without first being initialized
3523 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3524 warn if a non-static reference or non-static @samp{const} member
3525 appears in a class without constructors.
3527 If you want to warn about code which uses the uninitialized value of the
3528 variable in its own initializer, use the @option{-Winit-self} option.
3530 These warnings occur for individual uninitialized or clobbered
3531 elements of structure, union or array variables as well as for
3532 variables which are uninitialized or clobbered as a whole. They do
3533 not occur for variables or elements declared @code{volatile}. Because
3534 these warnings depend on optimization, the exact variables or elements
3535 for which there are warnings will depend on the precise optimization
3536 options and version of GCC used.
3538 Note that there may be no warning about a variable that is used only
3539 to compute a value that itself is never used, because such
3540 computations may be deleted by data flow analysis before the warnings
3543 These warnings are made optional because GCC is not smart
3544 enough to see all the reasons why the code might be correct
3545 despite appearing to have an error. Here is one example of how
3566 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3567 always initialized, but GCC doesn't know this. Here is
3568 another common case:
3573 if (change_y) save_y = y, y = new_y;
3575 if (change_y) y = save_y;
3580 This has no bug because @code{save_y} is used only if it is set.
3582 @cindex @code{longjmp} warnings
3583 This option also warns when a non-volatile automatic variable might be
3584 changed by a call to @code{longjmp}. These warnings as well are possible
3585 only in optimizing compilation.
3587 The compiler sees only the calls to @code{setjmp}. It cannot know
3588 where @code{longjmp} will be called; in fact, a signal handler could
3589 call it at any point in the code. As a result, you may get a warning
3590 even when there is in fact no problem because @code{longjmp} cannot
3591 in fact be called at the place which would cause a problem.
3593 Some spurious warnings can be avoided if you declare all the functions
3594 you use that never return as @code{noreturn}. @xref{Function
3597 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3599 @item -Wunknown-pragmas
3600 @opindex Wunknown-pragmas
3601 @opindex Wno-unknown-pragmas
3602 @cindex warning for unknown pragmas
3603 @cindex unknown pragmas, warning
3604 @cindex pragmas, warning of unknown
3605 Warn when a #pragma directive is encountered which is not understood by
3606 GCC@. If this command line option is used, warnings will even be issued
3607 for unknown pragmas in system header files. This is not the case if
3608 the warnings were only enabled by the @option{-Wall} command line option.
3611 @opindex Wno-pragmas
3613 Do not warn about misuses of pragmas, such as incorrect parameters,
3614 invalid syntax, or conflicts between pragmas. See also
3615 @samp{-Wunknown-pragmas}.
3617 @item -Wstrict-aliasing
3618 @opindex Wstrict-aliasing
3619 @opindex Wno-strict-aliasing
3620 This option is only active when @option{-fstrict-aliasing} is active.
3621 It warns about code which might break the strict aliasing rules that the
3622 compiler is using for optimization. The warning does not catch all
3623 cases, but does attempt to catch the more common pitfalls. It is
3624 included in @option{-Wall}.
3625 It is equivalent to @option{-Wstrict-aliasing=3}
3627 @item -Wstrict-aliasing=n
3628 @opindex Wstrict-aliasing=n
3629 @opindex Wno-strict-aliasing=n
3630 This option is only active when @option{-fstrict-aliasing} is active.
3631 It warns about code which might break the strict aliasing rules that the
3632 compiler is using for optimization.
3633 Higher levels correspond to higher accuracy (fewer false positives).
3634 Higher levels also correspond to more effort, similar to the way -O works.
3635 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3638 Level 1: Most aggressive, quick, least accurate.
3639 Possibly useful when higher levels
3640 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3641 false negatives. However, it has many false positives.
3642 Warns for all pointer conversions between possibly incompatible types,
3643 even if never dereferenced. Runs in the frontend only.
3645 Level 2: Aggressive, quick, not too precise.
3646 May still have many false positives (not as many as level 1 though),
3647 and few false negatives (but possibly more than level 1).
3648 Unlike level 1, it only warns when an address is taken. Warns about
3649 incomplete types. Runs in the frontend only.
3651 Level 3 (default for @option{-Wstrict-aliasing}):
3652 Should have very few false positives and few false
3653 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3654 Takes care of the common pun+dereference pattern in the frontend:
3655 @code{*(int*)&some_float}.
3656 If optimization is enabled, it also runs in the backend, where it deals
3657 with multiple statement cases using flow-sensitive points-to information.
3658 Only warns when the converted pointer is dereferenced.
3659 Does not warn about incomplete types.
3661 @item -Wstrict-overflow
3662 @itemx -Wstrict-overflow=@var{n}
3663 @opindex Wstrict-overflow
3664 @opindex Wno-strict-overflow
3665 This option is only active when @option{-fstrict-overflow} is active.
3666 It warns about cases where the compiler optimizes based on the
3667 assumption that signed overflow does not occur. Note that it does not
3668 warn about all cases where the code might overflow: it only warns
3669 about cases where the compiler implements some optimization. Thus
3670 this warning depends on the optimization level.
3672 An optimization which assumes that signed overflow does not occur is
3673 perfectly safe if the values of the variables involved are such that
3674 overflow never does, in fact, occur. Therefore this warning can
3675 easily give a false positive: a warning about code which is not
3676 actually a problem. To help focus on important issues, several
3677 warning levels are defined. No warnings are issued for the use of
3678 undefined signed overflow when estimating how many iterations a loop
3679 will require, in particular when determining whether a loop will be
3683 @item -Wstrict-overflow=1
3684 Warn about cases which are both questionable and easy to avoid. For
3685 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3686 compiler will simplify this to @code{1}. This level of
3687 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3688 are not, and must be explicitly requested.
3690 @item -Wstrict-overflow=2
3691 Also warn about other cases where a comparison is simplified to a
3692 constant. For example: @code{abs (x) >= 0}. This can only be
3693 simplified when @option{-fstrict-overflow} is in effect, because
3694 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3695 zero. @option{-Wstrict-overflow} (with no level) is the same as
3696 @option{-Wstrict-overflow=2}.
3698 @item -Wstrict-overflow=3
3699 Also warn about other cases where a comparison is simplified. For
3700 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3702 @item -Wstrict-overflow=4
3703 Also warn about other simplifications not covered by the above cases.
3704 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3706 @item -Wstrict-overflow=5
3707 Also warn about cases where the compiler reduces the magnitude of a
3708 constant involved in a comparison. For example: @code{x + 2 > y} will
3709 be simplified to @code{x + 1 >= y}. This is reported only at the
3710 highest warning level because this simplification applies to many
3711 comparisons, so this warning level will give a very large number of
3715 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3716 @opindex Wsuggest-attribute=
3717 @opindex Wno-suggest-attribute=
3718 Warn for cases where adding an attribute may be beneficial. The
3719 attributes currently supported are listed below.
3722 @item -Wsuggest-attribute=pure
3723 @itemx -Wsuggest-attribute=const
3724 @itemx -Wsuggest-attribute=noreturn
3725 @opindex Wsuggest-attribute=pure
3726 @opindex Wno-suggest-attribute=pure
3727 @opindex Wsuggest-attribute=const
3728 @opindex Wno-suggest-attribute=const
3729 @opindex Wsuggest-attribute=noreturn
3730 @opindex Wno-suggest-attribute=noreturn
3732 Warn about functions which might be candidates for attributes
3733 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3734 functions visible in other compilation units or (in the case of @code{pure} and
3735 @code{const}) if it cannot prove that the function returns normally. A function
3736 returns normally if it doesn't contain an infinite loop nor returns abnormally
3737 by throwing, calling @code{abort()} or trapping. This analysis requires option
3738 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3739 higher. Higher optimization levels improve the accuracy of the analysis.
3742 @item -Warray-bounds
3743 @opindex Wno-array-bounds
3744 @opindex Warray-bounds
3745 This option is only active when @option{-ftree-vrp} is active
3746 (default for @option{-O2} and above). It warns about subscripts to arrays
3747 that are always out of bounds. This warning is enabled by @option{-Wall}.
3749 @item -Wno-div-by-zero
3750 @opindex Wno-div-by-zero
3751 @opindex Wdiv-by-zero
3752 Do not warn about compile-time integer division by zero. Floating point
3753 division by zero is not warned about, as it can be a legitimate way of
3754 obtaining infinities and NaNs.
3756 @item -Wsystem-headers
3757 @opindex Wsystem-headers
3758 @opindex Wno-system-headers
3759 @cindex warnings from system headers
3760 @cindex system headers, warnings from
3761 Print warning messages for constructs found in system header files.
3762 Warnings from system headers are normally suppressed, on the assumption
3763 that they usually do not indicate real problems and would only make the
3764 compiler output harder to read. Using this command line option tells
3765 GCC to emit warnings from system headers as if they occurred in user
3766 code. However, note that using @option{-Wall} in conjunction with this
3767 option will @emph{not} warn about unknown pragmas in system
3768 headers---for that, @option{-Wunknown-pragmas} must also be used.
3771 @opindex Wtrampolines
3772 @opindex Wno-trampolines
3773 Warn about trampolines generated for pointers to nested functions.
3775 A trampoline is a small piece of data or code that is created at run
3776 time on the stack when the address of a nested function is taken, and
3777 is used to call the nested function indirectly. For some targets, it
3778 is made up of data only and thus requires no special treatment. But,
3779 for most targets, it is made up of code and thus requires the stack
3780 to be made executable in order for the program to work properly.
3783 @opindex Wfloat-equal
3784 @opindex Wno-float-equal
3785 Warn if floating point values are used in equality comparisons.
3787 The idea behind this is that sometimes it is convenient (for the
3788 programmer) to consider floating-point values as approximations to
3789 infinitely precise real numbers. If you are doing this, then you need
3790 to compute (by analyzing the code, or in some other way) the maximum or
3791 likely maximum error that the computation introduces, and allow for it
3792 when performing comparisons (and when producing output, but that's a
3793 different problem). In particular, instead of testing for equality, you
3794 would check to see whether the two values have ranges that overlap; and
3795 this is done with the relational operators, so equality comparisons are
3798 @item -Wtraditional @r{(C and Objective-C only)}
3799 @opindex Wtraditional
3800 @opindex Wno-traditional
3801 Warn about certain constructs that behave differently in traditional and
3802 ISO C@. Also warn about ISO C constructs that have no traditional C
3803 equivalent, and/or problematic constructs which should be avoided.
3807 Macro parameters that appear within string literals in the macro body.
3808 In traditional C macro replacement takes place within string literals,
3809 but does not in ISO C@.
3812 In traditional C, some preprocessor directives did not exist.
3813 Traditional preprocessors would only consider a line to be a directive
3814 if the @samp{#} appeared in column 1 on the line. Therefore
3815 @option{-Wtraditional} warns about directives that traditional C
3816 understands but would ignore because the @samp{#} does not appear as the
3817 first character on the line. It also suggests you hide directives like
3818 @samp{#pragma} not understood by traditional C by indenting them. Some
3819 traditional implementations would not recognize @samp{#elif}, so it
3820 suggests avoiding it altogether.
3823 A function-like macro that appears without arguments.
3826 The unary plus operator.
3829 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3830 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3831 constants.) Note, these suffixes appear in macros defined in the system
3832 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3833 Use of these macros in user code might normally lead to spurious
3834 warnings, however GCC's integrated preprocessor has enough context to
3835 avoid warning in these cases.
3838 A function declared external in one block and then used after the end of
3842 A @code{switch} statement has an operand of type @code{long}.
3845 A non-@code{static} function declaration follows a @code{static} one.
3846 This construct is not accepted by some traditional C compilers.
3849 The ISO type of an integer constant has a different width or
3850 signedness from its traditional type. This warning is only issued if
3851 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3852 typically represent bit patterns, are not warned about.
3855 Usage of ISO string concatenation is detected.
3858 Initialization of automatic aggregates.
3861 Identifier conflicts with labels. Traditional C lacks a separate
3862 namespace for labels.
3865 Initialization of unions. If the initializer is zero, the warning is
3866 omitted. This is done under the assumption that the zero initializer in
3867 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3868 initializer warnings and relies on default initialization to zero in the
3872 Conversions by prototypes between fixed/floating point values and vice
3873 versa. The absence of these prototypes when compiling with traditional
3874 C would cause serious problems. This is a subset of the possible
3875 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3878 Use of ISO C style function definitions. This warning intentionally is
3879 @emph{not} issued for prototype declarations or variadic functions
3880 because these ISO C features will appear in your code when using
3881 libiberty's traditional C compatibility macros, @code{PARAMS} and
3882 @code{VPARAMS}. This warning is also bypassed for nested functions
3883 because that feature is already a GCC extension and thus not relevant to
3884 traditional C compatibility.
3887 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3888 @opindex Wtraditional-conversion
3889 @opindex Wno-traditional-conversion
3890 Warn if a prototype causes a type conversion that is different from what
3891 would happen to the same argument in the absence of a prototype. This
3892 includes conversions of fixed point to floating and vice versa, and
3893 conversions changing the width or signedness of a fixed point argument
3894 except when the same as the default promotion.
3896 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3897 @opindex Wdeclaration-after-statement
3898 @opindex Wno-declaration-after-statement
3899 Warn when a declaration is found after a statement in a block. This
3900 construct, known from C++, was introduced with ISO C99 and is by default
3901 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3902 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3907 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3909 @item -Wno-endif-labels
3910 @opindex Wno-endif-labels
3911 @opindex Wendif-labels
3912 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3917 Warn whenever a local variable or type declaration shadows another variable,
3918 parameter, type, or class member (in C++), or whenever a built-in function
3919 is shadowed. Note that in C++, the compiler will not warn if a local variable
3920 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
3922 @item -Wlarger-than=@var{len}
3923 @opindex Wlarger-than=@var{len}
3924 @opindex Wlarger-than-@var{len}
3925 Warn whenever an object of larger than @var{len} bytes is defined.
3927 @item -Wframe-larger-than=@var{len}
3928 @opindex Wframe-larger-than
3929 Warn if the size of a function frame is larger than @var{len} bytes.
3930 The computation done to determine the stack frame size is approximate
3931 and not conservative.
3932 The actual requirements may be somewhat greater than @var{len}
3933 even if you do not get a warning. In addition, any space allocated
3934 via @code{alloca}, variable-length arrays, or related constructs
3935 is not included by the compiler when determining
3936 whether or not to issue a warning.
3938 @item -Wunsafe-loop-optimizations
3939 @opindex Wunsafe-loop-optimizations
3940 @opindex Wno-unsafe-loop-optimizations
3941 Warn if the loop cannot be optimized because the compiler could not
3942 assume anything on the bounds of the loop indices. With
3943 @option{-funsafe-loop-optimizations} warn if the compiler made
3946 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3947 @opindex Wno-pedantic-ms-format
3948 @opindex Wpedantic-ms-format
3949 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3950 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3951 depending on the MS runtime, when you are using the options @option{-Wformat}
3952 and @option{-pedantic} without gnu-extensions.
3954 @item -Wpointer-arith
3955 @opindex Wpointer-arith
3956 @opindex Wno-pointer-arith
3957 Warn about anything that depends on the ``size of'' a function type or
3958 of @code{void}. GNU C assigns these types a size of 1, for
3959 convenience in calculations with @code{void *} pointers and pointers
3960 to functions. In C++, warn also when an arithmetic operation involves
3961 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3964 @opindex Wtype-limits
3965 @opindex Wno-type-limits
3966 Warn if a comparison is always true or always false due to the limited
3967 range of the data type, but do not warn for constant expressions. For
3968 example, warn if an unsigned variable is compared against zero with
3969 @samp{<} or @samp{>=}. This warning is also enabled by
3972 @item -Wbad-function-cast @r{(C and Objective-C only)}
3973 @opindex Wbad-function-cast
3974 @opindex Wno-bad-function-cast
3975 Warn whenever a function call is cast to a non-matching type.
3976 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3978 @item -Wc++-compat @r{(C and Objective-C only)}
3979 Warn about ISO C constructs that are outside of the common subset of
3980 ISO C and ISO C++, e.g.@: request for implicit conversion from
3981 @code{void *} to a pointer to non-@code{void} type.
3983 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3984 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3985 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3986 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3990 @opindex Wno-cast-qual
3991 Warn whenever a pointer is cast so as to remove a type qualifier from
3992 the target type. For example, warn if a @code{const char *} is cast
3993 to an ordinary @code{char *}.
3995 Also warn when making a cast which introduces a type qualifier in an
3996 unsafe way. For example, casting @code{char **} to @code{const char **}
3997 is unsafe, as in this example:
4000 /* p is char ** value. */
4001 const char **q = (const char **) p;
4002 /* Assignment of readonly string to const char * is OK. */
4004 /* Now char** pointer points to read-only memory. */
4009 @opindex Wcast-align
4010 @opindex Wno-cast-align
4011 Warn whenever a pointer is cast such that the required alignment of the
4012 target is increased. For example, warn if a @code{char *} is cast to
4013 an @code{int *} on machines where integers can only be accessed at
4014 two- or four-byte boundaries.
4016 @item -Wwrite-strings
4017 @opindex Wwrite-strings
4018 @opindex Wno-write-strings
4019 When compiling C, give string constants the type @code{const
4020 char[@var{length}]} so that copying the address of one into a
4021 non-@code{const} @code{char *} pointer will get a warning. These
4022 warnings will help you find at compile time code that can try to write
4023 into a string constant, but only if you have been very careful about
4024 using @code{const} in declarations and prototypes. Otherwise, it will
4025 just be a nuisance. This is why we did not make @option{-Wall} request
4028 When compiling C++, warn about the deprecated conversion from string
4029 literals to @code{char *}. This warning is enabled by default for C++
4034 @opindex Wno-clobbered
4035 Warn for variables that might be changed by @samp{longjmp} or
4036 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4039 @opindex Wconversion
4040 @opindex Wno-conversion
4041 Warn for implicit conversions that may alter a value. This includes
4042 conversions between real and integer, like @code{abs (x)} when
4043 @code{x} is @code{double}; conversions between signed and unsigned,
4044 like @code{unsigned ui = -1}; and conversions to smaller types, like
4045 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4046 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4047 changed by the conversion like in @code{abs (2.0)}. Warnings about
4048 conversions between signed and unsigned integers can be disabled by
4049 using @option{-Wno-sign-conversion}.
4051 For C++, also warn for confusing overload resolution for user-defined
4052 conversions; and conversions that will never use a type conversion
4053 operator: conversions to @code{void}, the same type, a base class or a
4054 reference to them. Warnings about conversions between signed and
4055 unsigned integers are disabled by default in C++ unless
4056 @option{-Wsign-conversion} is explicitly enabled.
4058 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4059 @opindex Wconversion-null
4060 @opindex Wno-conversion-null
4061 Do not warn for conversions between @code{NULL} and non-pointer
4062 types. @option{-Wconversion-null} is enabled by default.
4065 @opindex Wempty-body
4066 @opindex Wno-empty-body
4067 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4068 while} statement. This warning is also enabled by @option{-Wextra}.
4070 @item -Wenum-compare
4071 @opindex Wenum-compare
4072 @opindex Wno-enum-compare
4073 Warn about a comparison between values of different enum types. In C++
4074 this warning is enabled by default. In C this warning is enabled by
4077 @item -Wjump-misses-init @r{(C, Objective-C only)}
4078 @opindex Wjump-misses-init
4079 @opindex Wno-jump-misses-init
4080 Warn if a @code{goto} statement or a @code{switch} statement jumps
4081 forward across the initialization of a variable, or jumps backward to a
4082 label after the variable has been initialized. This only warns about
4083 variables which are initialized when they are declared. This warning is
4084 only supported for C and Objective C; in C++ this sort of branch is an
4087 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4088 can be disabled with the @option{-Wno-jump-misses-init} option.
4090 @item -Wsign-compare
4091 @opindex Wsign-compare
4092 @opindex Wno-sign-compare
4093 @cindex warning for comparison of signed and unsigned values
4094 @cindex comparison of signed and unsigned values, warning
4095 @cindex signed and unsigned values, comparison warning
4096 Warn when a comparison between signed and unsigned values could produce
4097 an incorrect result when the signed value is converted to unsigned.
4098 This warning is also enabled by @option{-Wextra}; to get the other warnings
4099 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4101 @item -Wsign-conversion
4102 @opindex Wsign-conversion
4103 @opindex Wno-sign-conversion
4104 Warn for implicit conversions that may change the sign of an integer
4105 value, like assigning a signed integer expression to an unsigned
4106 integer variable. An explicit cast silences the warning. In C, this
4107 option is enabled also by @option{-Wconversion}.
4111 @opindex Wno-address
4112 Warn about suspicious uses of memory addresses. These include using
4113 the address of a function in a conditional expression, such as
4114 @code{void func(void); if (func)}, and comparisons against the memory
4115 address of a string literal, such as @code{if (x == "abc")}. Such
4116 uses typically indicate a programmer error: the address of a function
4117 always evaluates to true, so their use in a conditional usually
4118 indicate that the programmer forgot the parentheses in a function
4119 call; and comparisons against string literals result in unspecified
4120 behavior and are not portable in C, so they usually indicate that the
4121 programmer intended to use @code{strcmp}. This warning is enabled by
4125 @opindex Wlogical-op
4126 @opindex Wno-logical-op
4127 Warn about suspicious uses of logical operators in expressions.
4128 This includes using logical operators in contexts where a
4129 bit-wise operator is likely to be expected.
4131 @item -Waggregate-return
4132 @opindex Waggregate-return
4133 @opindex Wno-aggregate-return
4134 Warn if any functions that return structures or unions are defined or
4135 called. (In languages where you can return an array, this also elicits
4138 @item -Wno-attributes
4139 @opindex Wno-attributes
4140 @opindex Wattributes
4141 Do not warn if an unexpected @code{__attribute__} is used, such as
4142 unrecognized attributes, function attributes applied to variables,
4143 etc. This will not stop errors for incorrect use of supported
4146 @item -Wno-builtin-macro-redefined
4147 @opindex Wno-builtin-macro-redefined
4148 @opindex Wbuiltin-macro-redefined
4149 Do not warn if certain built-in macros are redefined. This suppresses
4150 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4151 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4153 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4154 @opindex Wstrict-prototypes
4155 @opindex Wno-strict-prototypes
4156 Warn if a function is declared or defined without specifying the
4157 argument types. (An old-style function definition is permitted without
4158 a warning if preceded by a declaration which specifies the argument
4161 @item -Wold-style-declaration @r{(C and Objective-C only)}
4162 @opindex Wold-style-declaration
4163 @opindex Wno-old-style-declaration
4164 Warn for obsolescent usages, according to the C Standard, in a
4165 declaration. For example, warn if storage-class specifiers like
4166 @code{static} are not the first things in a declaration. This warning
4167 is also enabled by @option{-Wextra}.
4169 @item -Wold-style-definition @r{(C and Objective-C only)}
4170 @opindex Wold-style-definition
4171 @opindex Wno-old-style-definition
4172 Warn if an old-style function definition is used. A warning is given
4173 even if there is a previous prototype.
4175 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4176 @opindex Wmissing-parameter-type
4177 @opindex Wno-missing-parameter-type
4178 A function parameter is declared without a type specifier in K&R-style
4185 This warning is also enabled by @option{-Wextra}.
4187 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4188 @opindex Wmissing-prototypes
4189 @opindex Wno-missing-prototypes
4190 Warn if a global function is defined without a previous prototype
4191 declaration. This warning is issued even if the definition itself
4192 provides a prototype. The aim is to detect global functions that fail
4193 to be declared in header files.
4195 @item -Wmissing-declarations
4196 @opindex Wmissing-declarations
4197 @opindex Wno-missing-declarations
4198 Warn if a global function is defined without a previous declaration.
4199 Do so even if the definition itself provides a prototype.
4200 Use this option to detect global functions that are not declared in
4201 header files. In C++, no warnings are issued for function templates,
4202 or for inline functions, or for functions in anonymous namespaces.
4204 @item -Wmissing-field-initializers
4205 @opindex Wmissing-field-initializers
4206 @opindex Wno-missing-field-initializers
4210 Warn if a structure's initializer has some fields missing. For
4211 example, the following code would cause such a warning, because
4212 @code{x.h} is implicitly zero:
4215 struct s @{ int f, g, h; @};
4216 struct s x = @{ 3, 4 @};
4219 This option does not warn about designated initializers, so the following
4220 modification would not trigger a warning:
4223 struct s @{ int f, g, h; @};
4224 struct s x = @{ .f = 3, .g = 4 @};
4227 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4228 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4230 @item -Wmissing-format-attribute
4231 @opindex Wmissing-format-attribute
4232 @opindex Wno-missing-format-attribute
4235 Warn about function pointers which might be candidates for @code{format}
4236 attributes. Note these are only possible candidates, not absolute ones.
4237 GCC will guess that function pointers with @code{format} attributes that
4238 are used in assignment, initialization, parameter passing or return
4239 statements should have a corresponding @code{format} attribute in the
4240 resulting type. I.e.@: the left-hand side of the assignment or
4241 initialization, the type of the parameter variable, or the return type
4242 of the containing function respectively should also have a @code{format}
4243 attribute to avoid the warning.
4245 GCC will also warn about function definitions which might be
4246 candidates for @code{format} attributes. Again, these are only
4247 possible candidates. GCC will guess that @code{format} attributes
4248 might be appropriate for any function that calls a function like
4249 @code{vprintf} or @code{vscanf}, but this might not always be the
4250 case, and some functions for which @code{format} attributes are
4251 appropriate may not be detected.
4253 @item -Wno-multichar
4254 @opindex Wno-multichar
4256 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4257 Usually they indicate a typo in the user's code, as they have
4258 implementation-defined values, and should not be used in portable code.
4260 @item -Wnormalized=<none|id|nfc|nfkc>
4261 @opindex Wnormalized=
4264 @cindex character set, input normalization
4265 In ISO C and ISO C++, two identifiers are different if they are
4266 different sequences of characters. However, sometimes when characters
4267 outside the basic ASCII character set are used, you can have two
4268 different character sequences that look the same. To avoid confusion,
4269 the ISO 10646 standard sets out some @dfn{normalization rules} which
4270 when applied ensure that two sequences that look the same are turned into
4271 the same sequence. GCC can warn you if you are using identifiers which
4272 have not been normalized; this option controls that warning.
4274 There are four levels of warning that GCC supports. The default is
4275 @option{-Wnormalized=nfc}, which warns about any identifier which is
4276 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4277 recommended form for most uses.
4279 Unfortunately, there are some characters which ISO C and ISO C++ allow
4280 in identifiers that when turned into NFC aren't allowable as
4281 identifiers. That is, there's no way to use these symbols in portable
4282 ISO C or C++ and have all your identifiers in NFC@.
4283 @option{-Wnormalized=id} suppresses the warning for these characters.
4284 It is hoped that future versions of the standards involved will correct
4285 this, which is why this option is not the default.
4287 You can switch the warning off for all characters by writing
4288 @option{-Wnormalized=none}. You would only want to do this if you
4289 were using some other normalization scheme (like ``D''), because
4290 otherwise you can easily create bugs that are literally impossible to see.
4292 Some characters in ISO 10646 have distinct meanings but look identical
4293 in some fonts or display methodologies, especially once formatting has
4294 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4295 LETTER N'', will display just like a regular @code{n} which has been
4296 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4297 normalization scheme to convert all these into a standard form as
4298 well, and GCC will warn if your code is not in NFKC if you use
4299 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4300 about every identifier that contains the letter O because it might be
4301 confused with the digit 0, and so is not the default, but may be
4302 useful as a local coding convention if the programming environment is
4303 unable to be fixed to display these characters distinctly.
4305 @item -Wno-deprecated
4306 @opindex Wno-deprecated
4307 @opindex Wdeprecated
4308 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4310 @item -Wno-deprecated-declarations
4311 @opindex Wno-deprecated-declarations
4312 @opindex Wdeprecated-declarations
4313 Do not warn about uses of functions (@pxref{Function Attributes}),
4314 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4315 Attributes}) marked as deprecated by using the @code{deprecated}
4319 @opindex Wno-overflow
4321 Do not warn about compile-time overflow in constant expressions.
4323 @item -Woverride-init @r{(C and Objective-C only)}
4324 @opindex Woverride-init
4325 @opindex Wno-override-init
4329 Warn if an initialized field without side effects is overridden when
4330 using designated initializers (@pxref{Designated Inits, , Designated
4333 This warning is included in @option{-Wextra}. To get other
4334 @option{-Wextra} warnings without this one, use @samp{-Wextra
4335 -Wno-override-init}.
4340 Warn if a structure is given the packed attribute, but the packed
4341 attribute has no effect on the layout or size of the structure.
4342 Such structures may be mis-aligned for little benefit. For
4343 instance, in this code, the variable @code{f.x} in @code{struct bar}
4344 will be misaligned even though @code{struct bar} does not itself
4345 have the packed attribute:
4352 @} __attribute__((packed));
4360 @item -Wpacked-bitfield-compat
4361 @opindex Wpacked-bitfield-compat
4362 @opindex Wno-packed-bitfield-compat
4363 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4364 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4365 the change can lead to differences in the structure layout. GCC
4366 informs you when the offset of such a field has changed in GCC 4.4.
4367 For example there is no longer a 4-bit padding between field @code{a}
4368 and @code{b} in this structure:
4375 @} __attribute__ ((packed));
4378 This warning is enabled by default. Use
4379 @option{-Wno-packed-bitfield-compat} to disable this warning.
4384 Warn if padding is included in a structure, either to align an element
4385 of the structure or to align the whole structure. Sometimes when this
4386 happens it is possible to rearrange the fields of the structure to
4387 reduce the padding and so make the structure smaller.
4389 @item -Wredundant-decls
4390 @opindex Wredundant-decls
4391 @opindex Wno-redundant-decls
4392 Warn if anything is declared more than once in the same scope, even in
4393 cases where multiple declaration is valid and changes nothing.
4395 @item -Wnested-externs @r{(C and Objective-C only)}
4396 @opindex Wnested-externs
4397 @opindex Wno-nested-externs
4398 Warn if an @code{extern} declaration is encountered within a function.
4403 Warn if a function can not be inlined and it was declared as inline.
4404 Even with this option, the compiler will not warn about failures to
4405 inline functions declared in system headers.
4407 The compiler uses a variety of heuristics to determine whether or not
4408 to inline a function. For example, the compiler takes into account
4409 the size of the function being inlined and the amount of inlining
4410 that has already been done in the current function. Therefore,
4411 seemingly insignificant changes in the source program can cause the
4412 warnings produced by @option{-Winline} to appear or disappear.
4414 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4415 @opindex Wno-invalid-offsetof
4416 @opindex Winvalid-offsetof
4417 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4418 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4419 to a non-POD type is undefined. In existing C++ implementations,
4420 however, @samp{offsetof} typically gives meaningful results even when
4421 applied to certain kinds of non-POD types. (Such as a simple
4422 @samp{struct} that fails to be a POD type only by virtue of having a
4423 constructor.) This flag is for users who are aware that they are
4424 writing nonportable code and who have deliberately chosen to ignore the
4427 The restrictions on @samp{offsetof} may be relaxed in a future version
4428 of the C++ standard.
4430 @item -Wno-int-to-pointer-cast
4431 @opindex Wno-int-to-pointer-cast
4432 @opindex Wint-to-pointer-cast
4433 Suppress warnings from casts to pointer type of an integer of a
4434 different size. In C++, casting to a pointer type of smaller size is
4435 an error. @option{Wint-to-pointer-cast} is enabled by default.
4438 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4439 @opindex Wno-pointer-to-int-cast
4440 @opindex Wpointer-to-int-cast
4441 Suppress warnings from casts from a pointer to an integer type of a
4445 @opindex Winvalid-pch
4446 @opindex Wno-invalid-pch
4447 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4448 the search path but can't be used.
4452 @opindex Wno-long-long
4453 Warn if @samp{long long} type is used. This is enabled by either
4454 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4455 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4457 @item -Wvariadic-macros
4458 @opindex Wvariadic-macros
4459 @opindex Wno-variadic-macros
4460 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4461 alternate syntax when in pedantic ISO C99 mode. This is default.
4462 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4467 Warn if variable length array is used in the code.
4468 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4469 the variable length array.
4471 @item -Wvolatile-register-var
4472 @opindex Wvolatile-register-var
4473 @opindex Wno-volatile-register-var
4474 Warn if a register variable is declared volatile. The volatile
4475 modifier does not inhibit all optimizations that may eliminate reads
4476 and/or writes to register variables. This warning is enabled by
4479 @item -Wdisabled-optimization
4480 @opindex Wdisabled-optimization
4481 @opindex Wno-disabled-optimization
4482 Warn if a requested optimization pass is disabled. This warning does
4483 not generally indicate that there is anything wrong with your code; it
4484 merely indicates that GCC's optimizers were unable to handle the code
4485 effectively. Often, the problem is that your code is too big or too
4486 complex; GCC will refuse to optimize programs when the optimization
4487 itself is likely to take inordinate amounts of time.
4489 @item -Wpointer-sign @r{(C and Objective-C only)}
4490 @opindex Wpointer-sign
4491 @opindex Wno-pointer-sign
4492 Warn for pointer argument passing or assignment with different signedness.
4493 This option is only supported for C and Objective-C@. It is implied by
4494 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4495 @option{-Wno-pointer-sign}.
4497 @item -Wstack-protector
4498 @opindex Wstack-protector
4499 @opindex Wno-stack-protector
4500 This option is only active when @option{-fstack-protector} is active. It
4501 warns about functions that will not be protected against stack smashing.
4504 @opindex Wno-mudflap
4505 Suppress warnings about constructs that cannot be instrumented by
4508 @item -Woverlength-strings
4509 @opindex Woverlength-strings
4510 @opindex Wno-overlength-strings
4511 Warn about string constants which are longer than the ``minimum
4512 maximum'' length specified in the C standard. Modern compilers
4513 generally allow string constants which are much longer than the
4514 standard's minimum limit, but very portable programs should avoid
4515 using longer strings.
4517 The limit applies @emph{after} string constant concatenation, and does
4518 not count the trailing NUL@. In C90, the limit was 509 characters; in
4519 C99, it was raised to 4095. C++98 does not specify a normative
4520 minimum maximum, so we do not diagnose overlength strings in C++@.
4522 This option is implied by @option{-pedantic}, and can be disabled with
4523 @option{-Wno-overlength-strings}.
4525 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4526 @opindex Wunsuffixed-float-constants
4528 GCC will issue a warning for any floating constant that does not have
4529 a suffix. When used together with @option{-Wsystem-headers} it will
4530 warn about such constants in system header files. This can be useful
4531 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4532 from the decimal floating-point extension to C99.
4535 @node Debugging Options
4536 @section Options for Debugging Your Program or GCC
4537 @cindex options, debugging
4538 @cindex debugging information options
4540 GCC has various special options that are used for debugging
4541 either your program or GCC:
4546 Produce debugging information in the operating system's native format
4547 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4550 On most systems that use stabs format, @option{-g} enables use of extra
4551 debugging information that only GDB can use; this extra information
4552 makes debugging work better in GDB but will probably make other debuggers
4554 refuse to read the program. If you want to control for certain whether
4555 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4556 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4558 GCC allows you to use @option{-g} with
4559 @option{-O}. The shortcuts taken by optimized code may occasionally
4560 produce surprising results: some variables you declared may not exist
4561 at all; flow of control may briefly move where you did not expect it;
4562 some statements may not be executed because they compute constant
4563 results or their values were already at hand; some statements may
4564 execute in different places because they were moved out of loops.
4566 Nevertheless it proves possible to debug optimized output. This makes
4567 it reasonable to use the optimizer for programs that might have bugs.
4569 The following options are useful when GCC is generated with the
4570 capability for more than one debugging format.
4574 Produce debugging information for use by GDB@. This means to use the
4575 most expressive format available (DWARF 2, stabs, or the native format
4576 if neither of those are supported), including GDB extensions if at all
4581 Produce debugging information in stabs format (if that is supported),
4582 without GDB extensions. This is the format used by DBX on most BSD
4583 systems. On MIPS, Alpha and System V Release 4 systems this option
4584 produces stabs debugging output which is not understood by DBX or SDB@.
4585 On System V Release 4 systems this option requires the GNU assembler.
4587 @item -feliminate-unused-debug-symbols
4588 @opindex feliminate-unused-debug-symbols
4589 Produce debugging information in stabs format (if that is supported),
4590 for only symbols that are actually used.
4592 @item -femit-class-debug-always
4593 Instead of emitting debugging information for a C++ class in only one
4594 object file, emit it in all object files using the class. This option
4595 should be used only with debuggers that are unable to handle the way GCC
4596 normally emits debugging information for classes because using this
4597 option will increase the size of debugging information by as much as a
4602 Produce debugging information in stabs format (if that is supported),
4603 using GNU extensions understood only by the GNU debugger (GDB)@. The
4604 use of these extensions is likely to make other debuggers crash or
4605 refuse to read the program.
4609 Produce debugging information in COFF format (if that is supported).
4610 This is the format used by SDB on most System V systems prior to
4615 Produce debugging information in XCOFF format (if that is supported).
4616 This is the format used by the DBX debugger on IBM RS/6000 systems.
4620 Produce debugging information in XCOFF format (if that is supported),
4621 using GNU extensions understood only by the GNU debugger (GDB)@. The
4622 use of these extensions is likely to make other debuggers crash or
4623 refuse to read the program, and may cause assemblers other than the GNU
4624 assembler (GAS) to fail with an error.
4626 @item -gdwarf-@var{version}
4627 @opindex gdwarf-@var{version}
4628 Produce debugging information in DWARF format (if that is
4629 supported). This is the format used by DBX on IRIX 6. The value
4630 of @var{version} may be either 2, 3 or 4; the default version is 2.
4632 Note that with DWARF version 2 some ports require, and will always
4633 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4635 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4636 for maximum benefit.
4638 @item -gstrict-dwarf
4639 @opindex gstrict-dwarf
4640 Disallow using extensions of later DWARF standard version than selected
4641 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4642 DWARF extensions from later standard versions is allowed.
4644 @item -gno-strict-dwarf
4645 @opindex gno-strict-dwarf
4646 Allow using extensions of later DWARF standard version than selected with
4647 @option{-gdwarf-@var{version}}.
4651 Produce debugging information in VMS debug format (if that is
4652 supported). This is the format used by DEBUG on VMS systems.
4655 @itemx -ggdb@var{level}
4656 @itemx -gstabs@var{level}
4657 @itemx -gcoff@var{level}
4658 @itemx -gxcoff@var{level}
4659 @itemx -gvms@var{level}
4660 Request debugging information and also use @var{level} to specify how
4661 much information. The default level is 2.
4663 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4666 Level 1 produces minimal information, enough for making backtraces in
4667 parts of the program that you don't plan to debug. This includes
4668 descriptions of functions and external variables, but no information
4669 about local variables and no line numbers.
4671 Level 3 includes extra information, such as all the macro definitions
4672 present in the program. Some debuggers support macro expansion when
4673 you use @option{-g3}.
4675 @option{-gdwarf-2} does not accept a concatenated debug level, because
4676 GCC used to support an option @option{-gdwarf} that meant to generate
4677 debug information in version 1 of the DWARF format (which is very
4678 different from version 2), and it would have been too confusing. That
4679 debug format is long obsolete, but the option cannot be changed now.
4680 Instead use an additional @option{-g@var{level}} option to change the
4681 debug level for DWARF.
4685 Turn off generation of debug info, if leaving out this option would have
4686 generated it, or turn it on at level 2 otherwise. The position of this
4687 argument in the command line does not matter, it takes effect after all
4688 other options are processed, and it does so only once, no matter how
4689 many times it is given. This is mainly intended to be used with
4690 @option{-fcompare-debug}.
4692 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4693 @opindex fdump-final-insns
4694 Dump the final internal representation (RTL) to @var{file}. If the
4695 optional argument is omitted (or if @var{file} is @code{.}), the name
4696 of the dump file will be determined by appending @code{.gkd} to the
4697 compilation output file name.
4699 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4700 @opindex fcompare-debug
4701 @opindex fno-compare-debug
4702 If no error occurs during compilation, run the compiler a second time,
4703 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4704 passed to the second compilation. Dump the final internal
4705 representation in both compilations, and print an error if they differ.
4707 If the equal sign is omitted, the default @option{-gtoggle} is used.
4709 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4710 and nonzero, implicitly enables @option{-fcompare-debug}. If
4711 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4712 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4715 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4716 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4717 of the final representation and the second compilation, preventing even
4718 @env{GCC_COMPARE_DEBUG} from taking effect.
4720 To verify full coverage during @option{-fcompare-debug} testing, set
4721 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4722 which GCC will reject as an invalid option in any actual compilation
4723 (rather than preprocessing, assembly or linking). To get just a
4724 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4725 not overridden} will do.
4727 @item -fcompare-debug-second
4728 @opindex fcompare-debug-second
4729 This option is implicitly passed to the compiler for the second
4730 compilation requested by @option{-fcompare-debug}, along with options to
4731 silence warnings, and omitting other options that would cause
4732 side-effect compiler outputs to files or to the standard output. Dump
4733 files and preserved temporary files are renamed so as to contain the
4734 @code{.gk} additional extension during the second compilation, to avoid
4735 overwriting those generated by the first.
4737 When this option is passed to the compiler driver, it causes the
4738 @emph{first} compilation to be skipped, which makes it useful for little
4739 other than debugging the compiler proper.
4741 @item -feliminate-dwarf2-dups
4742 @opindex feliminate-dwarf2-dups
4743 Compress DWARF2 debugging information by eliminating duplicated
4744 information about each symbol. This option only makes sense when
4745 generating DWARF2 debugging information with @option{-gdwarf-2}.
4747 @item -femit-struct-debug-baseonly
4748 Emit debug information for struct-like types
4749 only when the base name of the compilation source file
4750 matches the base name of file in which the struct was defined.
4752 This option substantially reduces the size of debugging information,
4753 but at significant potential loss in type information to the debugger.
4754 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4755 See @option{-femit-struct-debug-detailed} for more detailed control.
4757 This option works only with DWARF 2.
4759 @item -femit-struct-debug-reduced
4760 Emit debug information for struct-like types
4761 only when the base name of the compilation source file
4762 matches the base name of file in which the type was defined,
4763 unless the struct is a template or defined in a system header.
4765 This option significantly reduces the size of debugging information,
4766 with some potential loss in type information to the debugger.
4767 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4768 See @option{-femit-struct-debug-detailed} for more detailed control.
4770 This option works only with DWARF 2.
4772 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4773 Specify the struct-like types
4774 for which the compiler will generate debug information.
4775 The intent is to reduce duplicate struct debug information
4776 between different object files within the same program.
4778 This option is a detailed version of
4779 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4780 which will serve for most needs.
4782 A specification has the syntax@*
4783 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4785 The optional first word limits the specification to
4786 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4787 A struct type is used directly when it is the type of a variable, member.
4788 Indirect uses arise through pointers to structs.
4789 That is, when use of an incomplete struct would be legal, the use is indirect.
4791 @samp{struct one direct; struct two * indirect;}.
4793 The optional second word limits the specification to
4794 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4795 Generic structs are a bit complicated to explain.
4796 For C++, these are non-explicit specializations of template classes,
4797 or non-template classes within the above.
4798 Other programming languages have generics,
4799 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4801 The third word specifies the source files for those
4802 structs for which the compiler will emit debug information.
4803 The values @samp{none} and @samp{any} have the normal meaning.
4804 The value @samp{base} means that
4805 the base of name of the file in which the type declaration appears
4806 must match the base of the name of the main compilation file.
4807 In practice, this means that
4808 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4809 but types declared in other header will not.
4810 The value @samp{sys} means those types satisfying @samp{base}
4811 or declared in system or compiler headers.
4813 You may need to experiment to determine the best settings for your application.
4815 The default is @samp{-femit-struct-debug-detailed=all}.
4817 This option works only with DWARF 2.
4819 @item -fno-merge-debug-strings
4820 @opindex fmerge-debug-strings
4821 @opindex fno-merge-debug-strings
4822 Direct the linker to not merge together strings in the debugging
4823 information which are identical in different object files. Merging is
4824 not supported by all assemblers or linkers. Merging decreases the size
4825 of the debug information in the output file at the cost of increasing
4826 link processing time. Merging is enabled by default.
4828 @item -fdebug-prefix-map=@var{old}=@var{new}
4829 @opindex fdebug-prefix-map
4830 When compiling files in directory @file{@var{old}}, record debugging
4831 information describing them as in @file{@var{new}} instead.
4833 @item -fno-dwarf2-cfi-asm
4834 @opindex fdwarf2-cfi-asm
4835 @opindex fno-dwarf2-cfi-asm
4836 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4837 instead of using GAS @code{.cfi_*} directives.
4839 @cindex @command{prof}
4842 Generate extra code to write profile information suitable for the
4843 analysis program @command{prof}. You must use this option when compiling
4844 the source files you want data about, and you must also use it when
4847 @cindex @command{gprof}
4850 Generate extra code to write profile information suitable for the
4851 analysis program @command{gprof}. You must use this option when compiling
4852 the source files you want data about, and you must also use it when
4857 Makes the compiler print out each function name as it is compiled, and
4858 print some statistics about each pass when it finishes.
4861 @opindex ftime-report
4862 Makes the compiler print some statistics about the time consumed by each
4863 pass when it finishes.
4866 @opindex fmem-report
4867 Makes the compiler print some statistics about permanent memory
4868 allocation when it finishes.
4870 @item -fpre-ipa-mem-report
4871 @opindex fpre-ipa-mem-report
4872 @item -fpost-ipa-mem-report
4873 @opindex fpost-ipa-mem-report
4874 Makes the compiler print some statistics about permanent memory
4875 allocation before or after interprocedural optimization.
4878 @opindex fstack-usage
4879 Makes the compiler output stack usage information for the program, on a
4880 per-function basis. The filename for the dump is made by appending
4881 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
4882 the output file, if explicitly specified and it is not an executable,
4883 otherwise it is the basename of the source file. An entry is made up
4888 The name of the function.
4892 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
4895 The qualifier @code{static} means that the function manipulates the stack
4896 statically: a fixed number of bytes are allocated for the frame on function
4897 entry and released on function exit; no stack adjustments are otherwise made
4898 in the function. The second field is this fixed number of bytes.
4900 The qualifier @code{dynamic} means that the function manipulates the stack
4901 dynamically: in addition to the static allocation described above, stack
4902 adjustments are made in the body of the function, for example to push/pop
4903 arguments around function calls. If the qualifier @code{bounded} is also
4904 present, the amount of these adjustments is bounded at compile-time and
4905 the second field is an upper bound of the total amount of stack used by
4906 the function. If it is not present, the amount of these adjustments is
4907 not bounded at compile-time and the second field only represents the
4910 @item -fprofile-arcs
4911 @opindex fprofile-arcs
4912 Add code so that program flow @dfn{arcs} are instrumented. During
4913 execution the program records how many times each branch and call is
4914 executed and how many times it is taken or returns. When the compiled
4915 program exits it saves this data to a file called
4916 @file{@var{auxname}.gcda} for each source file. The data may be used for
4917 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4918 test coverage analysis (@option{-ftest-coverage}). Each object file's
4919 @var{auxname} is generated from the name of the output file, if
4920 explicitly specified and it is not the final executable, otherwise it is
4921 the basename of the source file. In both cases any suffix is removed
4922 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4923 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4924 @xref{Cross-profiling}.
4926 @cindex @command{gcov}
4930 This option is used to compile and link code instrumented for coverage
4931 analysis. The option is a synonym for @option{-fprofile-arcs}
4932 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4933 linking). See the documentation for those options for more details.
4938 Compile the source files with @option{-fprofile-arcs} plus optimization
4939 and code generation options. For test coverage analysis, use the
4940 additional @option{-ftest-coverage} option. You do not need to profile
4941 every source file in a program.
4944 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4945 (the latter implies the former).
4948 Run the program on a representative workload to generate the arc profile
4949 information. This may be repeated any number of times. You can run
4950 concurrent instances of your program, and provided that the file system
4951 supports locking, the data files will be correctly updated. Also
4952 @code{fork} calls are detected and correctly handled (double counting
4956 For profile-directed optimizations, compile the source files again with
4957 the same optimization and code generation options plus
4958 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4959 Control Optimization}).
4962 For test coverage analysis, use @command{gcov} to produce human readable
4963 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4964 @command{gcov} documentation for further information.
4968 With @option{-fprofile-arcs}, for each function of your program GCC
4969 creates a program flow graph, then finds a spanning tree for the graph.
4970 Only arcs that are not on the spanning tree have to be instrumented: the
4971 compiler adds code to count the number of times that these arcs are
4972 executed. When an arc is the only exit or only entrance to a block, the
4973 instrumentation code can be added to the block; otherwise, a new basic
4974 block must be created to hold the instrumentation code.
4977 @item -ftest-coverage
4978 @opindex ftest-coverage
4979 Produce a notes file that the @command{gcov} code-coverage utility
4980 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4981 show program coverage. Each source file's note file is called
4982 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4983 above for a description of @var{auxname} and instructions on how to
4984 generate test coverage data. Coverage data will match the source files
4985 more closely, if you do not optimize.
4987 @item -fdbg-cnt-list
4988 @opindex fdbg-cnt-list
4989 Print the name and the counter upper bound for all debug counters.
4991 @item -fdbg-cnt=@var{counter-value-list}
4993 Set the internal debug counter upper bound. @var{counter-value-list}
4994 is a comma-separated list of @var{name}:@var{value} pairs
4995 which sets the upper bound of each debug counter @var{name} to @var{value}.
4996 All debug counters have the initial upper bound of @var{UINT_MAX},
4997 thus dbg_cnt() returns true always unless the upper bound is set by this option.
4998 e.g. With -fdbg-cnt=dce:10,tail_call:0
4999 dbg_cnt(dce) will return true only for first 10 invocations
5000 and dbg_cnt(tail_call) will return false always.
5002 @item -d@var{letters}
5003 @itemx -fdump-rtl-@var{pass}
5005 Says to make debugging dumps during compilation at times specified by
5006 @var{letters}. This is used for debugging the RTL-based passes of the
5007 compiler. The file names for most of the dumps are made by appending
5008 a pass number and a word to the @var{dumpname}, and the files are
5009 created in the directory of the output file. Note that the pass
5010 number is computed statically as passes get registered into the pass
5011 manager. Thus the numbering is not related to the dynamic order of
5012 execution of passes. In particular, a pass installed by a plugin
5013 could have a number over 200 even if it executed quite early.
5014 @var{dumpname} is generated from the name of the output file, if
5015 explicitly specified and it is not an executable, otherwise it is the
5016 basename of the source file. These switches may have different effects
5017 when @option{-E} is used for preprocessing.
5019 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5020 @option{-d} option @var{letters}. Here are the possible
5021 letters for use in @var{pass} and @var{letters}, and their meanings:
5025 @item -fdump-rtl-alignments
5026 @opindex fdump-rtl-alignments
5027 Dump after branch alignments have been computed.
5029 @item -fdump-rtl-asmcons
5030 @opindex fdump-rtl-asmcons
5031 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5033 @item -fdump-rtl-auto_inc_dec
5034 @opindex fdump-rtl-auto_inc_dec
5035 Dump after auto-inc-dec discovery. This pass is only run on
5036 architectures that have auto inc or auto dec instructions.
5038 @item -fdump-rtl-barriers
5039 @opindex fdump-rtl-barriers
5040 Dump after cleaning up the barrier instructions.
5042 @item -fdump-rtl-bbpart
5043 @opindex fdump-rtl-bbpart
5044 Dump after partitioning hot and cold basic blocks.
5046 @item -fdump-rtl-bbro
5047 @opindex fdump-rtl-bbro
5048 Dump after block reordering.
5050 @item -fdump-rtl-btl1
5051 @itemx -fdump-rtl-btl2
5052 @opindex fdump-rtl-btl2
5053 @opindex fdump-rtl-btl2
5054 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5055 after the two branch
5056 target load optimization passes.
5058 @item -fdump-rtl-bypass
5059 @opindex fdump-rtl-bypass
5060 Dump after jump bypassing and control flow optimizations.
5062 @item -fdump-rtl-combine
5063 @opindex fdump-rtl-combine
5064 Dump after the RTL instruction combination pass.
5066 @item -fdump-rtl-compgotos
5067 @opindex fdump-rtl-compgotos
5068 Dump after duplicating the computed gotos.
5070 @item -fdump-rtl-ce1
5071 @itemx -fdump-rtl-ce2
5072 @itemx -fdump-rtl-ce3
5073 @opindex fdump-rtl-ce1
5074 @opindex fdump-rtl-ce2
5075 @opindex fdump-rtl-ce3
5076 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5077 @option{-fdump-rtl-ce3} enable dumping after the three
5078 if conversion passes.
5080 @itemx -fdump-rtl-cprop_hardreg
5081 @opindex fdump-rtl-cprop_hardreg
5082 Dump after hard register copy propagation.
5084 @itemx -fdump-rtl-csa
5085 @opindex fdump-rtl-csa
5086 Dump after combining stack adjustments.
5088 @item -fdump-rtl-cse1
5089 @itemx -fdump-rtl-cse2
5090 @opindex fdump-rtl-cse1
5091 @opindex fdump-rtl-cse2
5092 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5093 the two common sub-expression elimination passes.
5095 @itemx -fdump-rtl-dce
5096 @opindex fdump-rtl-dce
5097 Dump after the standalone dead code elimination passes.
5099 @itemx -fdump-rtl-dbr
5100 @opindex fdump-rtl-dbr
5101 Dump after delayed branch scheduling.
5103 @item -fdump-rtl-dce1
5104 @itemx -fdump-rtl-dce2
5105 @opindex fdump-rtl-dce1
5106 @opindex fdump-rtl-dce2
5107 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5108 the two dead store elimination passes.
5111 @opindex fdump-rtl-eh
5112 Dump after finalization of EH handling code.
5114 @item -fdump-rtl-eh_ranges
5115 @opindex fdump-rtl-eh_ranges
5116 Dump after conversion of EH handling range regions.
5118 @item -fdump-rtl-expand
5119 @opindex fdump-rtl-expand
5120 Dump after RTL generation.
5122 @item -fdump-rtl-fwprop1
5123 @itemx -fdump-rtl-fwprop2
5124 @opindex fdump-rtl-fwprop1
5125 @opindex fdump-rtl-fwprop2
5126 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5127 dumping after the two forward propagation passes.
5129 @item -fdump-rtl-gcse1
5130 @itemx -fdump-rtl-gcse2
5131 @opindex fdump-rtl-gcse1
5132 @opindex fdump-rtl-gcse2
5133 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5134 after global common subexpression elimination.
5136 @item -fdump-rtl-init-regs
5137 @opindex fdump-rtl-init-regs
5138 Dump after the initialization of the registers.
5140 @item -fdump-rtl-initvals
5141 @opindex fdump-rtl-initvals
5142 Dump after the computation of the initial value sets.
5144 @itemx -fdump-rtl-into_cfglayout
5145 @opindex fdump-rtl-into_cfglayout
5146 Dump after converting to cfglayout mode.
5148 @item -fdump-rtl-ira
5149 @opindex fdump-rtl-ira
5150 Dump after iterated register allocation.
5152 @item -fdump-rtl-jump
5153 @opindex fdump-rtl-jump
5154 Dump after the second jump optimization.
5156 @item -fdump-rtl-loop2
5157 @opindex fdump-rtl-loop2
5158 @option{-fdump-rtl-loop2} enables dumping after the rtl
5159 loop optimization passes.
5161 @item -fdump-rtl-mach
5162 @opindex fdump-rtl-mach
5163 Dump after performing the machine dependent reorganization pass, if that
5166 @item -fdump-rtl-mode_sw
5167 @opindex fdump-rtl-mode_sw
5168 Dump after removing redundant mode switches.
5170 @item -fdump-rtl-rnreg
5171 @opindex fdump-rtl-rnreg
5172 Dump after register renumbering.
5174 @itemx -fdump-rtl-outof_cfglayout
5175 @opindex fdump-rtl-outof_cfglayout
5176 Dump after converting from cfglayout mode.
5178 @item -fdump-rtl-peephole2
5179 @opindex fdump-rtl-peephole2
5180 Dump after the peephole pass.
5182 @item -fdump-rtl-postreload
5183 @opindex fdump-rtl-postreload
5184 Dump after post-reload optimizations.
5186 @itemx -fdump-rtl-pro_and_epilogue
5187 @opindex fdump-rtl-pro_and_epilogue
5188 Dump after generating the function pro and epilogues.
5190 @item -fdump-rtl-regmove
5191 @opindex fdump-rtl-regmove
5192 Dump after the register move pass.
5194 @item -fdump-rtl-sched1
5195 @itemx -fdump-rtl-sched2
5196 @opindex fdump-rtl-sched1
5197 @opindex fdump-rtl-sched2
5198 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5199 after the basic block scheduling passes.
5201 @item -fdump-rtl-see
5202 @opindex fdump-rtl-see
5203 Dump after sign extension elimination.
5205 @item -fdump-rtl-seqabstr
5206 @opindex fdump-rtl-seqabstr
5207 Dump after common sequence discovery.
5209 @item -fdump-rtl-shorten
5210 @opindex fdump-rtl-shorten
5211 Dump after shortening branches.
5213 @item -fdump-rtl-sibling
5214 @opindex fdump-rtl-sibling
5215 Dump after sibling call optimizations.
5217 @item -fdump-rtl-split1
5218 @itemx -fdump-rtl-split2
5219 @itemx -fdump-rtl-split3
5220 @itemx -fdump-rtl-split4
5221 @itemx -fdump-rtl-split5
5222 @opindex fdump-rtl-split1
5223 @opindex fdump-rtl-split2
5224 @opindex fdump-rtl-split3
5225 @opindex fdump-rtl-split4
5226 @opindex fdump-rtl-split5
5227 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5228 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5229 @option{-fdump-rtl-split5} enable dumping after five rounds of
5230 instruction splitting.
5232 @item -fdump-rtl-sms
5233 @opindex fdump-rtl-sms
5234 Dump after modulo scheduling. This pass is only run on some
5237 @item -fdump-rtl-stack
5238 @opindex fdump-rtl-stack
5239 Dump after conversion from GCC's "flat register file" registers to the
5240 x87's stack-like registers. This pass is only run on x86 variants.
5242 @item -fdump-rtl-subreg1
5243 @itemx -fdump-rtl-subreg2
5244 @opindex fdump-rtl-subreg1
5245 @opindex fdump-rtl-subreg2
5246 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5247 the two subreg expansion passes.
5249 @item -fdump-rtl-unshare
5250 @opindex fdump-rtl-unshare
5251 Dump after all rtl has been unshared.
5253 @item -fdump-rtl-vartrack
5254 @opindex fdump-rtl-vartrack
5255 Dump after variable tracking.
5257 @item -fdump-rtl-vregs
5258 @opindex fdump-rtl-vregs
5259 Dump after converting virtual registers to hard registers.
5261 @item -fdump-rtl-web
5262 @opindex fdump-rtl-web
5263 Dump after live range splitting.
5265 @item -fdump-rtl-regclass
5266 @itemx -fdump-rtl-subregs_of_mode_init
5267 @itemx -fdump-rtl-subregs_of_mode_finish
5268 @itemx -fdump-rtl-dfinit
5269 @itemx -fdump-rtl-dfinish
5270 @opindex fdump-rtl-regclass
5271 @opindex fdump-rtl-subregs_of_mode_init
5272 @opindex fdump-rtl-subregs_of_mode_finish
5273 @opindex fdump-rtl-dfinit
5274 @opindex fdump-rtl-dfinish
5275 These dumps are defined but always produce empty files.
5277 @item -fdump-rtl-all
5278 @opindex fdump-rtl-all
5279 Produce all the dumps listed above.
5283 Annotate the assembler output with miscellaneous debugging information.
5287 Dump all macro definitions, at the end of preprocessing, in addition to
5292 Produce a core dump whenever an error occurs.
5296 Print statistics on memory usage, at the end of the run, to
5301 Annotate the assembler output with a comment indicating which
5302 pattern and alternative was used. The length of each instruction is
5307 Dump the RTL in the assembler output as a comment before each instruction.
5308 Also turns on @option{-dp} annotation.
5312 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5313 dump a representation of the control flow graph suitable for viewing with VCG
5314 to @file{@var{file}.@var{pass}.vcg}.
5318 Just generate RTL for a function instead of compiling it. Usually used
5319 with @option{-fdump-rtl-expand}.
5323 @opindex fdump-noaddr
5324 When doing debugging dumps, suppress address output. This makes it more
5325 feasible to use diff on debugging dumps for compiler invocations with
5326 different compiler binaries and/or different
5327 text / bss / data / heap / stack / dso start locations.
5329 @item -fdump-unnumbered
5330 @opindex fdump-unnumbered
5331 When doing debugging dumps, suppress instruction numbers and address output.
5332 This makes it more feasible to use diff on debugging dumps for compiler
5333 invocations with different options, in particular with and without
5336 @item -fdump-unnumbered-links
5337 @opindex fdump-unnumbered-links
5338 When doing debugging dumps (see @option{-d} option above), suppress
5339 instruction numbers for the links to the previous and next instructions
5342 @item -fdump-translation-unit @r{(C++ only)}
5343 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5344 @opindex fdump-translation-unit
5345 Dump a representation of the tree structure for the entire translation
5346 unit to a file. The file name is made by appending @file{.tu} to the
5347 source file name, and the file is created in the same directory as the
5348 output file. If the @samp{-@var{options}} form is used, @var{options}
5349 controls the details of the dump as described for the
5350 @option{-fdump-tree} options.
5352 @item -fdump-class-hierarchy @r{(C++ only)}
5353 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5354 @opindex fdump-class-hierarchy
5355 Dump a representation of each class's hierarchy and virtual function
5356 table layout to a file. The file name is made by appending
5357 @file{.class} to the source file name, and the file is created in the
5358 same directory as the output file. If the @samp{-@var{options}} form
5359 is used, @var{options} controls the details of the dump as described
5360 for the @option{-fdump-tree} options.
5362 @item -fdump-ipa-@var{switch}
5364 Control the dumping at various stages of inter-procedural analysis
5365 language tree to a file. The file name is generated by appending a
5366 switch specific suffix to the source file name, and the file is created
5367 in the same directory as the output file. The following dumps are
5372 Enables all inter-procedural analysis dumps.
5375 Dumps information about call-graph optimization, unused function removal,
5376 and inlining decisions.
5379 Dump after function inlining.
5383 @item -fdump-statistics-@var{option}
5384 @opindex fdump-statistics
5385 Enable and control dumping of pass statistics in a separate file. The
5386 file name is generated by appending a suffix ending in
5387 @samp{.statistics} to the source file name, and the file is created in
5388 the same directory as the output file. If the @samp{-@var{option}}
5389 form is used, @samp{-stats} will cause counters to be summed over the
5390 whole compilation unit while @samp{-details} will dump every event as
5391 the passes generate them. The default with no option is to sum
5392 counters for each function compiled.
5394 @item -fdump-tree-@var{switch}
5395 @itemx -fdump-tree-@var{switch}-@var{options}
5397 Control the dumping at various stages of processing the intermediate
5398 language tree to a file. The file name is generated by appending a
5399 switch specific suffix to the source file name, and the file is
5400 created in the same directory as the output file. If the
5401 @samp{-@var{options}} form is used, @var{options} is a list of
5402 @samp{-} separated options that control the details of the dump. Not
5403 all options are applicable to all dumps, those which are not
5404 meaningful will be ignored. The following options are available
5408 Print the address of each node. Usually this is not meaningful as it
5409 changes according to the environment and source file. Its primary use
5410 is for tying up a dump file with a debug environment.
5412 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5413 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5414 use working backward from mangled names in the assembly file.
5416 Inhibit dumping of members of a scope or body of a function merely
5417 because that scope has been reached. Only dump such items when they
5418 are directly reachable by some other path. When dumping pretty-printed
5419 trees, this option inhibits dumping the bodies of control structures.
5421 Print a raw representation of the tree. By default, trees are
5422 pretty-printed into a C-like representation.
5424 Enable more detailed dumps (not honored by every dump option).
5426 Enable dumping various statistics about the pass (not honored by every dump
5429 Enable showing basic block boundaries (disabled in raw dumps).
5431 Enable showing virtual operands for every statement.
5433 Enable showing line numbers for statements.
5435 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5437 Enable showing the tree dump for each statement.
5439 Enable showing the EH region number holding each statement.
5441 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5442 and @option{lineno}.
5445 The following tree dumps are possible:
5449 @opindex fdump-tree-original
5450 Dump before any tree based optimization, to @file{@var{file}.original}.
5453 @opindex fdump-tree-optimized
5454 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5457 @opindex fdump-tree-gimple
5458 Dump each function before and after the gimplification pass to a file. The
5459 file name is made by appending @file{.gimple} to the source file name.
5462 @opindex fdump-tree-cfg
5463 Dump the control flow graph of each function to a file. The file name is
5464 made by appending @file{.cfg} to the source file name.
5467 @opindex fdump-tree-vcg
5468 Dump the control flow graph of each function to a file in VCG format. The
5469 file name is made by appending @file{.vcg} to the source file name. Note
5470 that if the file contains more than one function, the generated file cannot
5471 be used directly by VCG@. You will need to cut and paste each function's
5472 graph into its own separate file first.
5475 @opindex fdump-tree-ch
5476 Dump each function after copying loop headers. The file name is made by
5477 appending @file{.ch} to the source file name.
5480 @opindex fdump-tree-ssa
5481 Dump SSA related information to a file. The file name is made by appending
5482 @file{.ssa} to the source file name.
5485 @opindex fdump-tree-alias
5486 Dump aliasing information for each function. The file name is made by
5487 appending @file{.alias} to the source file name.
5490 @opindex fdump-tree-ccp
5491 Dump each function after CCP@. The file name is made by appending
5492 @file{.ccp} to the source file name.
5495 @opindex fdump-tree-storeccp
5496 Dump each function after STORE-CCP@. The file name is made by appending
5497 @file{.storeccp} to the source file name.
5500 @opindex fdump-tree-pre
5501 Dump trees after partial redundancy elimination. The file name is made
5502 by appending @file{.pre} to the source file name.
5505 @opindex fdump-tree-fre
5506 Dump trees after full redundancy elimination. The file name is made
5507 by appending @file{.fre} to the source file name.
5510 @opindex fdump-tree-copyprop
5511 Dump trees after copy propagation. The file name is made
5512 by appending @file{.copyprop} to the source file name.
5514 @item store_copyprop
5515 @opindex fdump-tree-store_copyprop
5516 Dump trees after store copy-propagation. The file name is made
5517 by appending @file{.store_copyprop} to the source file name.
5520 @opindex fdump-tree-dce
5521 Dump each function after dead code elimination. The file name is made by
5522 appending @file{.dce} to the source file name.
5525 @opindex fdump-tree-mudflap
5526 Dump each function after adding mudflap instrumentation. The file name is
5527 made by appending @file{.mudflap} to the source file name.
5530 @opindex fdump-tree-sra
5531 Dump each function after performing scalar replacement of aggregates. The
5532 file name is made by appending @file{.sra} to the source file name.
5535 @opindex fdump-tree-sink
5536 Dump each function after performing code sinking. The file name is made
5537 by appending @file{.sink} to the source file name.
5540 @opindex fdump-tree-dom
5541 Dump each function after applying dominator tree optimizations. The file
5542 name is made by appending @file{.dom} to the source file name.
5545 @opindex fdump-tree-dse
5546 Dump each function after applying dead store elimination. The file
5547 name is made by appending @file{.dse} to the source file name.
5550 @opindex fdump-tree-phiopt
5551 Dump each function after optimizing PHI nodes into straightline code. The file
5552 name is made by appending @file{.phiopt} to the source file name.
5555 @opindex fdump-tree-forwprop
5556 Dump each function after forward propagating single use variables. The file
5557 name is made by appending @file{.forwprop} to the source file name.
5560 @opindex fdump-tree-copyrename
5561 Dump each function after applying the copy rename optimization. The file
5562 name is made by appending @file{.copyrename} to the source file name.
5565 @opindex fdump-tree-nrv
5566 Dump each function after applying the named return value optimization on
5567 generic trees. The file name is made by appending @file{.nrv} to the source
5571 @opindex fdump-tree-vect
5572 Dump each function after applying vectorization of loops. The file name is
5573 made by appending @file{.vect} to the source file name.
5576 @opindex fdump-tree-slp
5577 Dump each function after applying vectorization of basic blocks. The file name
5578 is made by appending @file{.slp} to the source file name.
5581 @opindex fdump-tree-vrp
5582 Dump each function after Value Range Propagation (VRP). The file name
5583 is made by appending @file{.vrp} to the source file name.
5586 @opindex fdump-tree-all
5587 Enable all the available tree dumps with the flags provided in this option.
5590 @item -ftree-vectorizer-verbose=@var{n}
5591 @opindex ftree-vectorizer-verbose
5592 This option controls the amount of debugging output the vectorizer prints.
5593 This information is written to standard error, unless
5594 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5595 in which case it is output to the usual dump listing file, @file{.vect}.
5596 For @var{n}=0 no diagnostic information is reported.
5597 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5598 and the total number of loops that got vectorized.
5599 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5600 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5601 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5602 level that @option{-fdump-tree-vect-stats} uses.
5603 Higher verbosity levels mean either more information dumped for each
5604 reported loop, or same amount of information reported for more loops:
5605 if @var{n}=3, vectorizer cost model information is reported.
5606 If @var{n}=4, alignment related information is added to the reports.
5607 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5608 memory access-patterns) is added to the reports.
5609 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5610 that did not pass the first analysis phase (i.e., may not be countable, or
5611 may have complicated control-flow).
5612 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5613 If @var{n}=8, SLP related information is added to the reports.
5614 For @var{n}=9, all the information the vectorizer generates during its
5615 analysis and transformation is reported. This is the same verbosity level
5616 that @option{-fdump-tree-vect-details} uses.
5618 @item -frandom-seed=@var{string}
5619 @opindex frandom-seed
5620 This option provides a seed that GCC uses when it would otherwise use
5621 random numbers. It is used to generate certain symbol names
5622 that have to be different in every compiled file. It is also used to
5623 place unique stamps in coverage data files and the object files that
5624 produce them. You can use the @option{-frandom-seed} option to produce
5625 reproducibly identical object files.
5627 The @var{string} should be different for every file you compile.
5629 @item -fsched-verbose=@var{n}
5630 @opindex fsched-verbose
5631 On targets that use instruction scheduling, this option controls the
5632 amount of debugging output the scheduler prints. This information is
5633 written to standard error, unless @option{-fdump-rtl-sched1} or
5634 @option{-fdump-rtl-sched2} is specified, in which case it is output
5635 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5636 respectively. However for @var{n} greater than nine, the output is
5637 always printed to standard error.
5639 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5640 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5641 For @var{n} greater than one, it also output basic block probabilities,
5642 detailed ready list information and unit/insn info. For @var{n} greater
5643 than two, it includes RTL at abort point, control-flow and regions info.
5644 And for @var{n} over four, @option{-fsched-verbose} also includes
5648 @itemx -save-temps=cwd
5650 Store the usual ``temporary'' intermediate files permanently; place them
5651 in the current directory and name them based on the source file. Thus,
5652 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5653 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5654 preprocessed @file{foo.i} output file even though the compiler now
5655 normally uses an integrated preprocessor.
5657 When used in combination with the @option{-x} command line option,
5658 @option{-save-temps} is sensible enough to avoid over writing an
5659 input source file with the same extension as an intermediate file.
5660 The corresponding intermediate file may be obtained by renaming the
5661 source file before using @option{-save-temps}.
5663 If you invoke GCC in parallel, compiling several different source
5664 files that share a common base name in different subdirectories or the
5665 same source file compiled for multiple output destinations, it is
5666 likely that the different parallel compilers will interfere with each
5667 other, and overwrite the temporary files. For instance:
5670 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5671 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5674 may result in @file{foo.i} and @file{foo.o} being written to
5675 simultaneously by both compilers.
5677 @item -save-temps=obj
5678 @opindex save-temps=obj
5679 Store the usual ``temporary'' intermediate files permanently. If the
5680 @option{-o} option is used, the temporary files are based on the
5681 object file. If the @option{-o} option is not used, the
5682 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5687 gcc -save-temps=obj -c foo.c
5688 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5689 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5692 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5693 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5694 @file{dir2/yfoobar.o}.
5696 @item -time@r{[}=@var{file}@r{]}
5698 Report the CPU time taken by each subprocess in the compilation
5699 sequence. For C source files, this is the compiler proper and assembler
5700 (plus the linker if linking is done).
5702 Without the specification of an output file, the output looks like this:
5709 The first number on each line is the ``user time'', that is time spent
5710 executing the program itself. The second number is ``system time'',
5711 time spent executing operating system routines on behalf of the program.
5712 Both numbers are in seconds.
5714 With the specification of an output file, the output is appended to the
5715 named file, and it looks like this:
5718 0.12 0.01 cc1 @var{options}
5719 0.00 0.01 as @var{options}
5722 The ``user time'' and the ``system time'' are moved before the program
5723 name, and the options passed to the program are displayed, so that one
5724 can later tell what file was being compiled, and with which options.
5726 @item -fvar-tracking
5727 @opindex fvar-tracking
5728 Run variable tracking pass. It computes where variables are stored at each
5729 position in code. Better debugging information is then generated
5730 (if the debugging information format supports this information).
5732 It is enabled by default when compiling with optimization (@option{-Os},
5733 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5734 the debug info format supports it.
5736 @item -fvar-tracking-assignments
5737 @opindex fvar-tracking-assignments
5738 @opindex fno-var-tracking-assignments
5739 Annotate assignments to user variables early in the compilation and
5740 attempt to carry the annotations over throughout the compilation all the
5741 way to the end, in an attempt to improve debug information while
5742 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5744 It can be enabled even if var-tracking is disabled, in which case
5745 annotations will be created and maintained, but discarded at the end.
5747 @item -fvar-tracking-assignments-toggle
5748 @opindex fvar-tracking-assignments-toggle
5749 @opindex fno-var-tracking-assignments-toggle
5750 Toggle @option{-fvar-tracking-assignments}, in the same way that
5751 @option{-gtoggle} toggles @option{-g}.
5753 @item -print-file-name=@var{library}
5754 @opindex print-file-name
5755 Print the full absolute name of the library file @var{library} that
5756 would be used when linking---and don't do anything else. With this
5757 option, GCC does not compile or link anything; it just prints the
5760 @item -print-multi-directory
5761 @opindex print-multi-directory
5762 Print the directory name corresponding to the multilib selected by any
5763 other switches present in the command line. This directory is supposed
5764 to exist in @env{GCC_EXEC_PREFIX}.
5766 @item -print-multi-lib
5767 @opindex print-multi-lib
5768 Print the mapping from multilib directory names to compiler switches
5769 that enable them. The directory name is separated from the switches by
5770 @samp{;}, and each switch starts with an @samp{@@} instead of the
5771 @samp{-}, without spaces between multiple switches. This is supposed to
5772 ease shell-processing.
5774 @item -print-multi-os-directory
5775 @opindex print-multi-os-directory
5776 Print the path to OS libraries for the selected
5777 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5778 present in the @file{lib} subdirectory and no multilibs are used, this is
5779 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5780 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5781 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5782 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5784 @item -print-prog-name=@var{program}
5785 @opindex print-prog-name
5786 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5788 @item -print-libgcc-file-name
5789 @opindex print-libgcc-file-name
5790 Same as @option{-print-file-name=libgcc.a}.
5792 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5793 but you do want to link with @file{libgcc.a}. You can do
5796 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5799 @item -print-search-dirs
5800 @opindex print-search-dirs
5801 Print the name of the configured installation directory and a list of
5802 program and library directories @command{gcc} will search---and don't do anything else.
5804 This is useful when @command{gcc} prints the error message
5805 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5806 To resolve this you either need to put @file{cpp0} and the other compiler
5807 components where @command{gcc} expects to find them, or you can set the environment
5808 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5809 Don't forget the trailing @samp{/}.
5810 @xref{Environment Variables}.
5812 @item -print-sysroot
5813 @opindex print-sysroot
5814 Print the target sysroot directory that will be used during
5815 compilation. This is the target sysroot specified either at configure
5816 time or using the @option{--sysroot} option, possibly with an extra
5817 suffix that depends on compilation options. If no target sysroot is
5818 specified, the option prints nothing.
5820 @item -print-sysroot-headers-suffix
5821 @opindex print-sysroot-headers-suffix
5822 Print the suffix added to the target sysroot when searching for
5823 headers, or give an error if the compiler is not configured with such
5824 a suffix---and don't do anything else.
5827 @opindex dumpmachine
5828 Print the compiler's target machine (for example,
5829 @samp{i686-pc-linux-gnu})---and don't do anything else.
5832 @opindex dumpversion
5833 Print the compiler version (for example, @samp{3.0})---and don't do
5838 Print the compiler's built-in specs---and don't do anything else. (This
5839 is used when GCC itself is being built.) @xref{Spec Files}.
5841 @item -feliminate-unused-debug-types
5842 @opindex feliminate-unused-debug-types
5843 Normally, when producing DWARF2 output, GCC will emit debugging
5844 information for all types declared in a compilation
5845 unit, regardless of whether or not they are actually used
5846 in that compilation unit. Sometimes this is useful, such as
5847 if, in the debugger, you want to cast a value to a type that is
5848 not actually used in your program (but is declared). More often,
5849 however, this results in a significant amount of wasted space.
5850 With this option, GCC will avoid producing debug symbol output
5851 for types that are nowhere used in the source file being compiled.
5854 @node Optimize Options
5855 @section Options That Control Optimization
5856 @cindex optimize options
5857 @cindex options, optimization
5859 These options control various sorts of optimizations.
5861 Without any optimization option, the compiler's goal is to reduce the
5862 cost of compilation and to make debugging produce the expected
5863 results. Statements are independent: if you stop the program with a
5864 breakpoint between statements, you can then assign a new value to any
5865 variable or change the program counter to any other statement in the
5866 function and get exactly the results you would expect from the source
5869 Turning on optimization flags makes the compiler attempt to improve
5870 the performance and/or code size at the expense of compilation time
5871 and possibly the ability to debug the program.
5873 The compiler performs optimization based on the knowledge it has of the
5874 program. Compiling multiple files at once to a single output file mode allows
5875 the compiler to use information gained from all of the files when compiling
5878 Not all optimizations are controlled directly by a flag. Only
5879 optimizations that have a flag are listed in this section.
5881 Most optimizations are only enabled if an @option{-O} level is set on
5882 the command line. Otherwise they are disabled, even if individual
5883 optimization flags are specified.
5885 Depending on the target and how GCC was configured, a slightly different
5886 set of optimizations may be enabled at each @option{-O} level than
5887 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5888 to find out the exact set of optimizations that are enabled at each level.
5889 @xref{Overall Options}, for examples.
5896 Optimize. Optimizing compilation takes somewhat more time, and a lot
5897 more memory for a large function.
5899 With @option{-O}, the compiler tries to reduce code size and execution
5900 time, without performing any optimizations that take a great deal of
5903 @option{-O} turns on the following optimization flags:
5907 -fcprop-registers @gol
5910 -fdelayed-branch @gol
5912 -fguess-branch-probability @gol
5913 -fif-conversion2 @gol
5914 -fif-conversion @gol
5915 -fipa-pure-const @gol
5917 -fipa-reference @gol
5919 -fsplit-wide-types @gol
5921 -ftree-builtin-call-dce @gol
5924 -ftree-copyrename @gol
5926 -ftree-dominator-opts @gol
5928 -ftree-forwprop @gol
5936 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5937 where doing so does not interfere with debugging.
5941 Optimize even more. GCC performs nearly all supported optimizations
5942 that do not involve a space-speed tradeoff.
5943 As compared to @option{-O}, this option increases both compilation time
5944 and the performance of the generated code.
5946 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5947 also turns on the following optimization flags:
5948 @gccoptlist{-fthread-jumps @gol
5949 -falign-functions -falign-jumps @gol
5950 -falign-loops -falign-labels @gol
5953 -fcse-follow-jumps -fcse-skip-blocks @gol
5954 -fdelete-null-pointer-checks @gol
5956 -fexpensive-optimizations @gol
5957 -fgcse -fgcse-lm @gol
5958 -finline-small-functions @gol
5959 -findirect-inlining @gol
5961 -foptimize-sibling-calls @gol
5962 -fpartial-inlining @gol
5965 -freorder-blocks -freorder-functions @gol
5966 -frerun-cse-after-loop @gol
5967 -fsched-interblock -fsched-spec @gol
5968 -fschedule-insns -fschedule-insns2 @gol
5969 -fstrict-aliasing -fstrict-overflow @gol
5970 -ftree-switch-conversion @gol
5974 Please note the warning under @option{-fgcse} about
5975 invoking @option{-O2} on programs that use computed gotos.
5979 Optimize yet more. @option{-O3} turns on all optimizations specified
5980 by @option{-O2} and also turns on the @option{-finline-functions},
5981 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5982 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
5983 @option{-fipa-cp-clone} options.
5987 Reduce compilation time and make debugging produce the expected
5988 results. This is the default.
5992 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5993 do not typically increase code size. It also performs further
5994 optimizations designed to reduce code size.
5996 @option{-Os} disables the following optimization flags:
5997 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5998 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5999 -fprefetch-loop-arrays -ftree-vect-loop-version}
6003 Disregard strict standards compliance. @option{-Ofast} enables all
6004 @option{-O3} optimizations. It also enables optimizations that are not
6005 valid for all standard compliant programs.
6006 It turns on @option{-ffast-math}.
6008 If you use multiple @option{-O} options, with or without level numbers,
6009 the last such option is the one that is effective.
6012 Options of the form @option{-f@var{flag}} specify machine-independent
6013 flags. Most flags have both positive and negative forms; the negative
6014 form of @option{-ffoo} would be @option{-fno-foo}. In the table
6015 below, only one of the forms is listed---the one you typically will
6016 use. You can figure out the other form by either removing @samp{no-}
6019 The following options control specific optimizations. They are either
6020 activated by @option{-O} options or are related to ones that are. You
6021 can use the following flags in the rare cases when ``fine-tuning'' of
6022 optimizations to be performed is desired.
6025 @item -fno-default-inline
6026 @opindex fno-default-inline
6027 Do not make member functions inline by default merely because they are
6028 defined inside the class scope (C++ only). Otherwise, when you specify
6029 @w{@option{-O}}, member functions defined inside class scope are compiled
6030 inline by default; i.e., you don't need to add @samp{inline} in front of
6031 the member function name.
6033 @item -fno-defer-pop
6034 @opindex fno-defer-pop
6035 Always pop the arguments to each function call as soon as that function
6036 returns. For machines which must pop arguments after a function call,
6037 the compiler normally lets arguments accumulate on the stack for several
6038 function calls and pops them all at once.
6040 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6042 @item -fforward-propagate
6043 @opindex fforward-propagate
6044 Perform a forward propagation pass on RTL@. The pass tries to combine two
6045 instructions and checks if the result can be simplified. If loop unrolling
6046 is active, two passes are performed and the second is scheduled after
6049 This option is enabled by default at optimization levels @option{-O},
6050 @option{-O2}, @option{-O3}, @option{-Os}.
6052 @item -ffp-contract=@var{style}
6053 @opindex ffp-contract
6054 @option{-ffp-contract=off} disables floating-point expression contraction.
6055 @option{-ffp-contract=fast} enables floating-point expression contraction
6056 such as forming of fused multiply-add operations if the target has
6057 native support for them.
6058 @option{-ffp-contract=on} enables floating-point expression contraction
6059 if allowed by the language standard. This is currently not implemented
6060 and treated equal to @option{-ffp-contract=off}.
6062 The default is @option{-ffp-contract=fast}.
6064 @item -fomit-frame-pointer
6065 @opindex fomit-frame-pointer
6066 Don't keep the frame pointer in a register for functions that
6067 don't need one. This avoids the instructions to save, set up and
6068 restore frame pointers; it also makes an extra register available
6069 in many functions. @strong{It also makes debugging impossible on
6072 On some machines, such as the VAX, this flag has no effect, because
6073 the standard calling sequence automatically handles the frame pointer
6074 and nothing is saved by pretending it doesn't exist. The
6075 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6076 whether a target machine supports this flag. @xref{Registers,,Register
6077 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6079 Starting with GCC version 4.6, the default setting (when not optimizing for
6080 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6081 @option{-fomit-frame-pointer}. The default can be reverted to
6082 @option{-fno-omit-frame-pointer} by configuring GCC with the
6083 @option{--enable-frame-pointer} configure option.
6085 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6087 @item -foptimize-sibling-calls
6088 @opindex foptimize-sibling-calls
6089 Optimize sibling and tail recursive calls.
6091 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6095 Don't pay attention to the @code{inline} keyword. Normally this option
6096 is used to keep the compiler from expanding any functions inline.
6097 Note that if you are not optimizing, no functions can be expanded inline.
6099 @item -finline-small-functions
6100 @opindex finline-small-functions
6101 Integrate functions into their callers when their body is smaller than expected
6102 function call code (so overall size of program gets smaller). The compiler
6103 heuristically decides which functions are simple enough to be worth integrating
6106 Enabled at level @option{-O2}.
6108 @item -findirect-inlining
6109 @opindex findirect-inlining
6110 Inline also indirect calls that are discovered to be known at compile
6111 time thanks to previous inlining. This option has any effect only
6112 when inlining itself is turned on by the @option{-finline-functions}
6113 or @option{-finline-small-functions} options.
6115 Enabled at level @option{-O2}.
6117 @item -finline-functions
6118 @opindex finline-functions
6119 Integrate all simple functions into their callers. The compiler
6120 heuristically decides which functions are simple enough to be worth
6121 integrating in this way.
6123 If all calls to a given function are integrated, and the function is
6124 declared @code{static}, then the function is normally not output as
6125 assembler code in its own right.
6127 Enabled at level @option{-O3}.
6129 @item -finline-functions-called-once
6130 @opindex finline-functions-called-once
6131 Consider all @code{static} functions called once for inlining into their
6132 caller even if they are not marked @code{inline}. If a call to a given
6133 function is integrated, then the function is not output as assembler code
6136 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6138 @item -fearly-inlining
6139 @opindex fearly-inlining
6140 Inline functions marked by @code{always_inline} and functions whose body seems
6141 smaller than the function call overhead early before doing
6142 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6143 makes profiling significantly cheaper and usually inlining faster on programs
6144 having large chains of nested wrapper functions.
6150 Perform interprocedural scalar replacement of aggregates, removal of
6151 unused parameters and replacement of parameters passed by reference
6152 by parameters passed by value.
6154 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6156 @item -finline-limit=@var{n}
6157 @opindex finline-limit
6158 By default, GCC limits the size of functions that can be inlined. This flag
6159 allows coarse control of this limit. @var{n} is the size of functions that
6160 can be inlined in number of pseudo instructions.
6162 Inlining is actually controlled by a number of parameters, which may be
6163 specified individually by using @option{--param @var{name}=@var{value}}.
6164 The @option{-finline-limit=@var{n}} option sets some of these parameters
6168 @item max-inline-insns-single
6169 is set to @var{n}/2.
6170 @item max-inline-insns-auto
6171 is set to @var{n}/2.
6174 See below for a documentation of the individual
6175 parameters controlling inlining and for the defaults of these parameters.
6177 @emph{Note:} there may be no value to @option{-finline-limit} that results
6178 in default behavior.
6180 @emph{Note:} pseudo instruction represents, in this particular context, an
6181 abstract measurement of function's size. In no way does it represent a count
6182 of assembly instructions and as such its exact meaning might change from one
6183 release to an another.
6185 @item -fno-keep-inline-dllexport
6186 @opindex -fno-keep-inline-dllexport
6187 This is a more fine-grained version of @option{-fkeep-inline-functions},
6188 which applies only to functions that are declared using the @code{dllexport}
6189 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6192 @item -fkeep-inline-functions
6193 @opindex fkeep-inline-functions
6194 In C, emit @code{static} functions that are declared @code{inline}
6195 into the object file, even if the function has been inlined into all
6196 of its callers. This switch does not affect functions using the
6197 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6198 inline functions into the object file.
6200 @item -fkeep-static-consts
6201 @opindex fkeep-static-consts
6202 Emit variables declared @code{static const} when optimization isn't turned
6203 on, even if the variables aren't referenced.
6205 GCC enables this option by default. If you want to force the compiler to
6206 check if the variable was referenced, regardless of whether or not
6207 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6209 @item -fmerge-constants
6210 @opindex fmerge-constants
6211 Attempt to merge identical constants (string constants and floating point
6212 constants) across compilation units.
6214 This option is the default for optimized compilation if the assembler and
6215 linker support it. Use @option{-fno-merge-constants} to inhibit this
6218 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6220 @item -fmerge-all-constants
6221 @opindex fmerge-all-constants
6222 Attempt to merge identical constants and identical variables.
6224 This option implies @option{-fmerge-constants}. In addition to
6225 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6226 arrays or initialized constant variables with integral or floating point
6227 types. Languages like C or C++ require each variable, including multiple
6228 instances of the same variable in recursive calls, to have distinct locations,
6229 so using this option will result in non-conforming
6232 @item -fmodulo-sched
6233 @opindex fmodulo-sched
6234 Perform swing modulo scheduling immediately before the first scheduling
6235 pass. This pass looks at innermost loops and reorders their
6236 instructions by overlapping different iterations.
6238 @item -fmodulo-sched-allow-regmoves
6239 @opindex fmodulo-sched-allow-regmoves
6240 Perform more aggressive SMS based modulo scheduling with register moves
6241 allowed. By setting this flag certain anti-dependences edges will be
6242 deleted which will trigger the generation of reg-moves based on the
6243 life-range analysis. This option is effective only with
6244 @option{-fmodulo-sched} enabled.
6246 @item -fno-branch-count-reg
6247 @opindex fno-branch-count-reg
6248 Do not use ``decrement and branch'' instructions on a count register,
6249 but instead generate a sequence of instructions that decrement a
6250 register, compare it against zero, then branch based upon the result.
6251 This option is only meaningful on architectures that support such
6252 instructions, which include x86, PowerPC, IA-64 and S/390.
6254 The default is @option{-fbranch-count-reg}.
6256 @item -fno-function-cse
6257 @opindex fno-function-cse
6258 Do not put function addresses in registers; make each instruction that
6259 calls a constant function contain the function's address explicitly.
6261 This option results in less efficient code, but some strange hacks
6262 that alter the assembler output may be confused by the optimizations
6263 performed when this option is not used.
6265 The default is @option{-ffunction-cse}
6267 @item -fno-zero-initialized-in-bss
6268 @opindex fno-zero-initialized-in-bss
6269 If the target supports a BSS section, GCC by default puts variables that
6270 are initialized to zero into BSS@. This can save space in the resulting
6273 This option turns off this behavior because some programs explicitly
6274 rely on variables going to the data section. E.g., so that the
6275 resulting executable can find the beginning of that section and/or make
6276 assumptions based on that.
6278 The default is @option{-fzero-initialized-in-bss}.
6280 @item -fmudflap -fmudflapth -fmudflapir
6284 @cindex bounds checking
6286 For front-ends that support it (C and C++), instrument all risky
6287 pointer/array dereferencing operations, some standard library
6288 string/heap functions, and some other associated constructs with
6289 range/validity tests. Modules so instrumented should be immune to
6290 buffer overflows, invalid heap use, and some other classes of C/C++
6291 programming errors. The instrumentation relies on a separate runtime
6292 library (@file{libmudflap}), which will be linked into a program if
6293 @option{-fmudflap} is given at link time. Run-time behavior of the
6294 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6295 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6298 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6299 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6300 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6301 instrumentation should ignore pointer reads. This produces less
6302 instrumentation (and therefore faster execution) and still provides
6303 some protection against outright memory corrupting writes, but allows
6304 erroneously read data to propagate within a program.
6306 @item -fthread-jumps
6307 @opindex fthread-jumps
6308 Perform optimizations where we check to see if a jump branches to a
6309 location where another comparison subsumed by the first is found. If
6310 so, the first branch is redirected to either the destination of the
6311 second branch or a point immediately following it, depending on whether
6312 the condition is known to be true or false.
6314 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6316 @item -fsplit-wide-types
6317 @opindex fsplit-wide-types
6318 When using a type that occupies multiple registers, such as @code{long
6319 long} on a 32-bit system, split the registers apart and allocate them
6320 independently. This normally generates better code for those types,
6321 but may make debugging more difficult.
6323 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6326 @item -fcse-follow-jumps
6327 @opindex fcse-follow-jumps
6328 In common subexpression elimination (CSE), scan through jump instructions
6329 when the target of the jump is not reached by any other path. For
6330 example, when CSE encounters an @code{if} statement with an
6331 @code{else} clause, CSE will follow the jump when the condition
6334 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6336 @item -fcse-skip-blocks
6337 @opindex fcse-skip-blocks
6338 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6339 follow jumps which conditionally skip over blocks. When CSE
6340 encounters a simple @code{if} statement with no else clause,
6341 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6342 body of the @code{if}.
6344 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6346 @item -frerun-cse-after-loop
6347 @opindex frerun-cse-after-loop
6348 Re-run common subexpression elimination after loop optimizations has been
6351 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6355 Perform a global common subexpression elimination pass.
6356 This pass also performs global constant and copy propagation.
6358 @emph{Note:} When compiling a program using computed gotos, a GCC
6359 extension, you may get better runtime performance if you disable
6360 the global common subexpression elimination pass by adding
6361 @option{-fno-gcse} to the command line.
6363 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6367 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6368 attempt to move loads which are only killed by stores into themselves. This
6369 allows a loop containing a load/store sequence to be changed to a load outside
6370 the loop, and a copy/store within the loop.
6372 Enabled by default when gcse is enabled.
6376 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6377 global common subexpression elimination. This pass will attempt to move
6378 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6379 loops containing a load/store sequence can be changed to a load before
6380 the loop and a store after the loop.
6382 Not enabled at any optimization level.
6386 When @option{-fgcse-las} is enabled, the global common subexpression
6387 elimination pass eliminates redundant loads that come after stores to the
6388 same memory location (both partial and full redundancies).
6390 Not enabled at any optimization level.
6392 @item -fgcse-after-reload
6393 @opindex fgcse-after-reload
6394 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6395 pass is performed after reload. The purpose of this pass is to cleanup
6398 @item -funsafe-loop-optimizations
6399 @opindex funsafe-loop-optimizations
6400 If given, the loop optimizer will assume that loop indices do not
6401 overflow, and that the loops with nontrivial exit condition are not
6402 infinite. This enables a wider range of loop optimizations even if
6403 the loop optimizer itself cannot prove that these assumptions are valid.
6404 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6405 if it finds this kind of loop.
6407 @item -fcrossjumping
6408 @opindex fcrossjumping
6409 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6410 resulting code may or may not perform better than without cross-jumping.
6412 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6414 @item -fauto-inc-dec
6415 @opindex fauto-inc-dec
6416 Combine increments or decrements of addresses with memory accesses.
6417 This pass is always skipped on architectures that do not have
6418 instructions to support this. Enabled by default at @option{-O} and
6419 higher on architectures that support this.
6423 Perform dead code elimination (DCE) on RTL@.
6424 Enabled by default at @option{-O} and higher.
6428 Perform dead store elimination (DSE) on RTL@.
6429 Enabled by default at @option{-O} and higher.
6431 @item -fif-conversion
6432 @opindex fif-conversion
6433 Attempt to transform conditional jumps into branch-less equivalents. This
6434 include use of conditional moves, min, max, set flags and abs instructions, and
6435 some tricks doable by standard arithmetics. The use of conditional execution
6436 on chips where it is available is controlled by @code{if-conversion2}.
6438 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6440 @item -fif-conversion2
6441 @opindex fif-conversion2
6442 Use conditional execution (where available) to transform conditional jumps into
6443 branch-less equivalents.
6445 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6447 @item -fdelete-null-pointer-checks
6448 @opindex fdelete-null-pointer-checks
6449 Assume that programs cannot safely dereference null pointers, and that
6450 no code or data element resides there. This enables simple constant
6451 folding optimizations at all optimization levels. In addition, other
6452 optimization passes in GCC use this flag to control global dataflow
6453 analyses that eliminate useless checks for null pointers; these assume
6454 that if a pointer is checked after it has already been dereferenced,
6457 Note however that in some environments this assumption is not true.
6458 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6459 for programs which depend on that behavior.
6461 Some targets, especially embedded ones, disable this option at all levels.
6462 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6463 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6464 are enabled independently at different optimization levels.
6466 @item -fdevirtualize
6467 @opindex fdevirtualize
6468 Attempt to convert calls to virtual functions to direct calls. This
6469 is done both within a procedure and interprocedurally as part of
6470 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
6471 propagation (@option{-fipa-cp}).
6472 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6474 @item -fexpensive-optimizations
6475 @opindex fexpensive-optimizations
6476 Perform a number of minor optimizations that are relatively expensive.
6478 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6480 @item -foptimize-register-move
6482 @opindex foptimize-register-move
6484 Attempt to reassign register numbers in move instructions and as
6485 operands of other simple instructions in order to maximize the amount of
6486 register tying. This is especially helpful on machines with two-operand
6489 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6492 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6494 @item -fira-algorithm=@var{algorithm}
6495 Use specified coloring algorithm for the integrated register
6496 allocator. The @var{algorithm} argument should be @code{priority} or
6497 @code{CB}. The first algorithm specifies Chow's priority coloring,
6498 the second one specifies Chaitin-Briggs coloring. The second
6499 algorithm can be unimplemented for some architectures. If it is
6500 implemented, it is the default because Chaitin-Briggs coloring as a
6501 rule generates a better code.
6503 @item -fira-region=@var{region}
6504 Use specified regions for the integrated register allocator. The
6505 @var{region} argument should be one of @code{all}, @code{mixed}, or
6506 @code{one}. The first value means using all loops as register
6507 allocation regions, the second value which is the default means using
6508 all loops except for loops with small register pressure as the
6509 regions, and third one means using all function as a single region.
6510 The first value can give best result for machines with small size and
6511 irregular register set, the third one results in faster and generates
6512 decent code and the smallest size code, and the default value usually
6513 give the best results in most cases and for most architectures.
6515 @item -fira-loop-pressure
6516 @opindex fira-loop-pressure
6517 Use IRA to evaluate register pressure in loops for decision to move
6518 loop invariants. Usage of this option usually results in generation
6519 of faster and smaller code on machines with big register files (>= 32
6520 registers) but it can slow compiler down.
6522 This option is enabled at level @option{-O3} for some targets.
6524 @item -fno-ira-share-save-slots
6525 @opindex fno-ira-share-save-slots
6526 Switch off sharing stack slots used for saving call used hard
6527 registers living through a call. Each hard register will get a
6528 separate stack slot and as a result function stack frame will be
6531 @item -fno-ira-share-spill-slots
6532 @opindex fno-ira-share-spill-slots
6533 Switch off sharing stack slots allocated for pseudo-registers. Each
6534 pseudo-register which did not get a hard register will get a separate
6535 stack slot and as a result function stack frame will be bigger.
6537 @item -fira-verbose=@var{n}
6538 @opindex fira-verbose
6539 Set up how verbose dump file for the integrated register allocator
6540 will be. Default value is 5. If the value is greater or equal to 10,
6541 the dump file will be stderr as if the value were @var{n} minus 10.
6543 @item -fdelayed-branch
6544 @opindex fdelayed-branch
6545 If supported for the target machine, attempt to reorder instructions
6546 to exploit instruction slots available after delayed branch
6549 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6551 @item -fschedule-insns
6552 @opindex fschedule-insns
6553 If supported for the target machine, attempt to reorder instructions to
6554 eliminate execution stalls due to required data being unavailable. This
6555 helps machines that have slow floating point or memory load instructions
6556 by allowing other instructions to be issued until the result of the load
6557 or floating point instruction is required.
6559 Enabled at levels @option{-O2}, @option{-O3}.
6561 @item -fschedule-insns2
6562 @opindex fschedule-insns2
6563 Similar to @option{-fschedule-insns}, but requests an additional pass of
6564 instruction scheduling after register allocation has been done. This is
6565 especially useful on machines with a relatively small number of
6566 registers and where memory load instructions take more than one cycle.
6568 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6570 @item -fno-sched-interblock
6571 @opindex fno-sched-interblock
6572 Don't schedule instructions across basic blocks. This is normally
6573 enabled by default when scheduling before register allocation, i.e.@:
6574 with @option{-fschedule-insns} or at @option{-O2} or higher.
6576 @item -fno-sched-spec
6577 @opindex fno-sched-spec
6578 Don't allow speculative motion of non-load instructions. This is normally
6579 enabled by default when scheduling before register allocation, i.e.@:
6580 with @option{-fschedule-insns} or at @option{-O2} or higher.
6582 @item -fsched-pressure
6583 @opindex fsched-pressure
6584 Enable register pressure sensitive insn scheduling before the register
6585 allocation. This only makes sense when scheduling before register
6586 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6587 @option{-O2} or higher. Usage of this option can improve the
6588 generated code and decrease its size by preventing register pressure
6589 increase above the number of available hard registers and as a
6590 consequence register spills in the register allocation.
6592 @item -fsched-spec-load
6593 @opindex fsched-spec-load
6594 Allow speculative motion of some load instructions. This only makes
6595 sense when scheduling before register allocation, i.e.@: with
6596 @option{-fschedule-insns} or at @option{-O2} or higher.
6598 @item -fsched-spec-load-dangerous
6599 @opindex fsched-spec-load-dangerous
6600 Allow speculative motion of more load instructions. This only makes
6601 sense when scheduling before register allocation, i.e.@: with
6602 @option{-fschedule-insns} or at @option{-O2} or higher.
6604 @item -fsched-stalled-insns
6605 @itemx -fsched-stalled-insns=@var{n}
6606 @opindex fsched-stalled-insns
6607 Define how many insns (if any) can be moved prematurely from the queue
6608 of stalled insns into the ready list, during the second scheduling pass.
6609 @option{-fno-sched-stalled-insns} means that no insns will be moved
6610 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6611 on how many queued insns can be moved prematurely.
6612 @option{-fsched-stalled-insns} without a value is equivalent to
6613 @option{-fsched-stalled-insns=1}.
6615 @item -fsched-stalled-insns-dep
6616 @itemx -fsched-stalled-insns-dep=@var{n}
6617 @opindex fsched-stalled-insns-dep
6618 Define how many insn groups (cycles) will be examined for a dependency
6619 on a stalled insn that is candidate for premature removal from the queue
6620 of stalled insns. This has an effect only during the second scheduling pass,
6621 and only if @option{-fsched-stalled-insns} is used.
6622 @option{-fno-sched-stalled-insns-dep} is equivalent to
6623 @option{-fsched-stalled-insns-dep=0}.
6624 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6625 @option{-fsched-stalled-insns-dep=1}.
6627 @item -fsched2-use-superblocks
6628 @opindex fsched2-use-superblocks
6629 When scheduling after register allocation, do use superblock scheduling
6630 algorithm. Superblock scheduling allows motion across basic block boundaries
6631 resulting on faster schedules. This option is experimental, as not all machine
6632 descriptions used by GCC model the CPU closely enough to avoid unreliable
6633 results from the algorithm.
6635 This only makes sense when scheduling after register allocation, i.e.@: with
6636 @option{-fschedule-insns2} or at @option{-O2} or higher.
6638 @item -fsched-group-heuristic
6639 @opindex fsched-group-heuristic
6640 Enable the group heuristic in the scheduler. This heuristic favors
6641 the instruction that belongs to a schedule group. This is enabled
6642 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6643 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6645 @item -fsched-critical-path-heuristic
6646 @opindex fsched-critical-path-heuristic
6647 Enable the critical-path heuristic in the scheduler. This heuristic favors
6648 instructions on the critical path. This is enabled by default when
6649 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6650 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6652 @item -fsched-spec-insn-heuristic
6653 @opindex fsched-spec-insn-heuristic
6654 Enable the speculative instruction heuristic in the scheduler. This
6655 heuristic favors speculative instructions with greater dependency weakness.
6656 This is enabled by default when scheduling is enabled, i.e.@:
6657 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6658 or at @option{-O2} or higher.
6660 @item -fsched-rank-heuristic
6661 @opindex fsched-rank-heuristic
6662 Enable the rank heuristic in the scheduler. This heuristic favors
6663 the instruction belonging to a basic block with greater size or frequency.
6664 This is enabled by default when scheduling is enabled, i.e.@:
6665 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6666 at @option{-O2} or higher.
6668 @item -fsched-last-insn-heuristic
6669 @opindex fsched-last-insn-heuristic
6670 Enable the last-instruction heuristic in the scheduler. This heuristic
6671 favors the instruction that is less dependent on the last instruction
6672 scheduled. This is enabled by default when scheduling is enabled,
6673 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6674 at @option{-O2} or higher.
6676 @item -fsched-dep-count-heuristic
6677 @opindex fsched-dep-count-heuristic
6678 Enable the dependent-count heuristic in the scheduler. This heuristic
6679 favors the instruction that has more instructions depending on it.
6680 This is enabled by default when scheduling is enabled, i.e.@:
6681 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6682 at @option{-O2} or higher.
6684 @item -freschedule-modulo-scheduled-loops
6685 @opindex freschedule-modulo-scheduled-loops
6686 The modulo scheduling comes before the traditional scheduling, if a loop
6687 was modulo scheduled we may want to prevent the later scheduling passes
6688 from changing its schedule, we use this option to control that.
6690 @item -fselective-scheduling
6691 @opindex fselective-scheduling
6692 Schedule instructions using selective scheduling algorithm. Selective
6693 scheduling runs instead of the first scheduler pass.
6695 @item -fselective-scheduling2
6696 @opindex fselective-scheduling2
6697 Schedule instructions using selective scheduling algorithm. Selective
6698 scheduling runs instead of the second scheduler pass.
6700 @item -fsel-sched-pipelining
6701 @opindex fsel-sched-pipelining
6702 Enable software pipelining of innermost loops during selective scheduling.
6703 This option has no effect until one of @option{-fselective-scheduling} or
6704 @option{-fselective-scheduling2} is turned on.
6706 @item -fsel-sched-pipelining-outer-loops
6707 @opindex fsel-sched-pipelining-outer-loops
6708 When pipelining loops during selective scheduling, also pipeline outer loops.
6709 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6711 @item -fcaller-saves
6712 @opindex fcaller-saves
6713 Enable values to be allocated in registers that will be clobbered by
6714 function calls, by emitting extra instructions to save and restore the
6715 registers around such calls. Such allocation is done only when it
6716 seems to result in better code than would otherwise be produced.
6718 This option is always enabled by default on certain machines, usually
6719 those which have no call-preserved registers to use instead.
6721 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6723 @item -fcombine-stack-adjustments
6724 @opindex fcombine-stack-adjustments
6725 Tracks stack adjustments (pushes and pops) and stack memory references
6726 and then tries to find ways to combine them.
6728 Enabled by default at @option{-O1} and higher.
6730 @item -fconserve-stack
6731 @opindex fconserve-stack
6732 Attempt to minimize stack usage. The compiler will attempt to use less
6733 stack space, even if that makes the program slower. This option
6734 implies setting the @option{large-stack-frame} parameter to 100
6735 and the @option{large-stack-frame-growth} parameter to 400.
6737 @item -ftree-reassoc
6738 @opindex ftree-reassoc
6739 Perform reassociation on trees. This flag is enabled by default
6740 at @option{-O} and higher.
6744 Perform partial redundancy elimination (PRE) on trees. This flag is
6745 enabled by default at @option{-O2} and @option{-O3}.
6747 @item -ftree-forwprop
6748 @opindex ftree-forwprop
6749 Perform forward propagation on trees. This flag is enabled by default
6750 at @option{-O} and higher.
6754 Perform full redundancy elimination (FRE) on trees. The difference
6755 between FRE and PRE is that FRE only considers expressions
6756 that are computed on all paths leading to the redundant computation.
6757 This analysis is faster than PRE, though it exposes fewer redundancies.
6758 This flag is enabled by default at @option{-O} and higher.
6760 @item -ftree-phiprop
6761 @opindex ftree-phiprop
6762 Perform hoisting of loads from conditional pointers on trees. This
6763 pass is enabled by default at @option{-O} and higher.
6765 @item -ftree-copy-prop
6766 @opindex ftree-copy-prop
6767 Perform copy propagation on trees. This pass eliminates unnecessary
6768 copy operations. This flag is enabled by default at @option{-O} and
6771 @item -fipa-pure-const
6772 @opindex fipa-pure-const
6773 Discover which functions are pure or constant.
6774 Enabled by default at @option{-O} and higher.
6776 @item -fipa-reference
6777 @opindex fipa-reference
6778 Discover which static variables do not escape cannot escape the
6780 Enabled by default at @option{-O} and higher.
6782 @item -fipa-struct-reorg
6783 @opindex fipa-struct-reorg
6784 Perform structure reorganization optimization, that change C-like structures
6785 layout in order to better utilize spatial locality. This transformation is
6786 affective for programs containing arrays of structures. Available in two
6787 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6788 or static (which uses built-in heuristics). It works only in whole program
6789 mode, so it requires @option{-fwhole-program} to be
6790 enabled. Structures considered @samp{cold} by this transformation are not
6791 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6793 With this flag, the program debug info reflects a new structure layout.
6797 Perform interprocedural pointer analysis and interprocedural modification
6798 and reference analysis. This option can cause excessive memory and
6799 compile-time usage on large compilation units. It is not enabled by
6800 default at any optimization level.
6803 @opindex fipa-profile
6804 Perform interprocedural profile propagation. The functions called only from
6805 cold functions are marked as cold. Also functions executed once (such as
6806 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
6807 functions and loop less parts of functions executed once are then optimized for
6809 Enabled by default at @option{-O} and higher.
6813 Perform interprocedural constant propagation.
6814 This optimization analyzes the program to determine when values passed
6815 to functions are constants and then optimizes accordingly.
6816 This optimization can substantially increase performance
6817 if the application has constants passed to functions.
6818 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6820 @item -fipa-cp-clone
6821 @opindex fipa-cp-clone
6822 Perform function cloning to make interprocedural constant propagation stronger.
6823 When enabled, interprocedural constant propagation will perform function cloning
6824 when externally visible function can be called with constant arguments.
6825 Because this optimization can create multiple copies of functions,
6826 it may significantly increase code size
6827 (see @option{--param ipcp-unit-growth=@var{value}}).
6828 This flag is enabled by default at @option{-O3}.
6830 @item -fipa-matrix-reorg
6831 @opindex fipa-matrix-reorg
6832 Perform matrix flattening and transposing.
6833 Matrix flattening tries to replace an @math{m}-dimensional matrix
6834 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6835 This reduces the level of indirection needed for accessing the elements
6836 of the matrix. The second optimization is matrix transposing that
6837 attempts to change the order of the matrix's dimensions in order to
6838 improve cache locality.
6839 Both optimizations need the @option{-fwhole-program} flag.
6840 Transposing is enabled only if profiling information is available.
6844 Perform forward store motion on trees. This flag is
6845 enabled by default at @option{-O} and higher.
6847 @item -ftree-bit-ccp
6848 @opindex ftree-bit-ccp
6849 Perform sparse conditional bit constant propagation on trees and propagate
6850 pointer alignment information.
6851 This pass only operates on local scalar variables and is enabled by default
6852 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
6856 Perform sparse conditional constant propagation (CCP) on trees. This
6857 pass only operates on local scalar variables and is enabled by default
6858 at @option{-O} and higher.
6860 @item -ftree-switch-conversion
6861 Perform conversion of simple initializations in a switch to
6862 initializations from a scalar array. This flag is enabled by default
6863 at @option{-O2} and higher.
6867 Perform dead code elimination (DCE) on trees. This flag is enabled by
6868 default at @option{-O} and higher.
6870 @item -ftree-builtin-call-dce
6871 @opindex ftree-builtin-call-dce
6872 Perform conditional dead code elimination (DCE) for calls to builtin functions
6873 that may set @code{errno} but are otherwise side-effect free. This flag is
6874 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6877 @item -ftree-dominator-opts
6878 @opindex ftree-dominator-opts
6879 Perform a variety of simple scalar cleanups (constant/copy
6880 propagation, redundancy elimination, range propagation and expression
6881 simplification) based on a dominator tree traversal. This also
6882 performs jump threading (to reduce jumps to jumps). This flag is
6883 enabled by default at @option{-O} and higher.
6887 Perform dead store elimination (DSE) on trees. A dead store is a store into
6888 a memory location which will later be overwritten by another store without
6889 any intervening loads. In this case the earlier store can be deleted. This
6890 flag is enabled by default at @option{-O} and higher.
6894 Perform loop header copying on trees. This is beneficial since it increases
6895 effectiveness of code motion optimizations. It also saves one jump. This flag
6896 is enabled by default at @option{-O} and higher. It is not enabled
6897 for @option{-Os}, since it usually increases code size.
6899 @item -ftree-loop-optimize
6900 @opindex ftree-loop-optimize
6901 Perform loop optimizations on trees. This flag is enabled by default
6902 at @option{-O} and higher.
6904 @item -ftree-loop-linear
6905 @opindex ftree-loop-linear
6906 Perform loop interchange transformations on tree. Same as
6907 @option{-floop-interchange}. To use this code transformation, GCC has
6908 to be configured with @option{--with-ppl} and @option{--with-cloog} to
6909 enable the Graphite loop transformation infrastructure.
6911 @item -floop-interchange
6912 @opindex floop-interchange
6913 Perform loop interchange transformations on loops. Interchanging two
6914 nested loops switches the inner and outer loops. For example, given a
6919 A(J, I) = A(J, I) * C
6923 loop interchange will transform the loop as if the user had written:
6927 A(J, I) = A(J, I) * C
6931 which can be beneficial when @code{N} is larger than the caches,
6932 because in Fortran, the elements of an array are stored in memory
6933 contiguously by column, and the original loop iterates over rows,
6934 potentially creating at each access a cache miss. This optimization
6935 applies to all the languages supported by GCC and is not limited to
6936 Fortran. To use this code transformation, GCC has to be configured
6937 with @option{--with-ppl} and @option{--with-cloog} to enable the
6938 Graphite loop transformation infrastructure.
6940 @item -floop-strip-mine
6941 @opindex floop-strip-mine
6942 Perform loop strip mining transformations on loops. Strip mining
6943 splits a loop into two nested loops. The outer loop has strides
6944 equal to the strip size and the inner loop has strides of the
6945 original loop within a strip. The strip length can be changed
6946 using the @option{loop-block-tile-size} parameter. For example,
6953 loop strip mining will transform the loop as if the user had written:
6956 DO I = II, min (II + 50, N)
6961 This optimization applies to all the languages supported by GCC and is
6962 not limited to Fortran. To use this code transformation, GCC has to
6963 be configured with @option{--with-ppl} and @option{--with-cloog} to
6964 enable the Graphite loop transformation infrastructure.
6967 @opindex floop-block
6968 Perform loop blocking transformations on loops. Blocking strip mines
6969 each loop in the loop nest such that the memory accesses of the
6970 element loops fit inside caches. The strip length can be changed
6971 using the @option{loop-block-tile-size} parameter. For example, given
6976 A(J, I) = B(I) + C(J)
6980 loop blocking will transform the loop as if the user had written:
6984 DO I = II, min (II + 50, N)
6985 DO J = JJ, min (JJ + 50, M)
6986 A(J, I) = B(I) + C(J)
6992 which can be beneficial when @code{M} is larger than the caches,
6993 because the innermost loop will iterate over a smaller amount of data
6994 that can be kept in the caches. This optimization applies to all the
6995 languages supported by GCC and is not limited to Fortran. To use this
6996 code transformation, GCC has to be configured with @option{--with-ppl}
6997 and @option{--with-cloog} to enable the Graphite loop transformation
7000 @item -fgraphite-identity
7001 @opindex fgraphite-identity
7002 Enable the identity transformation for graphite. For every SCoP we generate
7003 the polyhedral representation and transform it back to gimple. Using
7004 @option{-fgraphite-identity} we can check the costs or benefits of the
7005 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7006 are also performed by the code generator CLooG, like index splitting and
7007 dead code elimination in loops.
7009 @item -floop-flatten
7010 @opindex floop-flatten
7011 Removes the loop nesting structure: transforms the loop nest into a
7012 single loop. This transformation can be useful to vectorize all the
7013 levels of the loop nest.
7015 @item -floop-parallelize-all
7016 @opindex floop-parallelize-all
7017 Use the Graphite data dependence analysis to identify loops that can
7018 be parallelized. Parallelize all the loops that can be analyzed to
7019 not contain loop carried dependences without checking that it is
7020 profitable to parallelize the loops.
7022 @item -fcheck-data-deps
7023 @opindex fcheck-data-deps
7024 Compare the results of several data dependence analyzers. This option
7025 is used for debugging the data dependence analyzers.
7027 @item -ftree-loop-if-convert
7028 Attempt to transform conditional jumps in the innermost loops to
7029 branch-less equivalents. The intent is to remove control-flow from
7030 the innermost loops in order to improve the ability of the
7031 vectorization pass to handle these loops. This is enabled by default
7032 if vectorization is enabled.
7034 @item -ftree-loop-if-convert-stores
7035 Attempt to also if-convert conditional jumps containing memory writes.
7036 This transformation can be unsafe for multi-threaded programs as it
7037 transforms conditional memory writes into unconditional memory writes.
7040 for (i = 0; i < N; i++)
7044 would be transformed to
7046 for (i = 0; i < N; i++)
7047 A[i] = cond ? expr : A[i];
7049 potentially producing data races.
7051 @item -ftree-loop-distribution
7052 Perform loop distribution. This flag can improve cache performance on
7053 big loop bodies and allow further loop optimizations, like
7054 parallelization or vectorization, to take place. For example, the loop
7071 @item -ftree-loop-distribute-patterns
7072 Perform loop distribution of patterns that can be code generated with
7073 calls to a library. This flag is enabled by default at @option{-O3}.
7075 This pass distributes the initialization loops and generates a call to
7076 memset zero. For example, the loop
7092 and the initialization loop is transformed into a call to memset zero.
7094 @item -ftree-loop-im
7095 @opindex ftree-loop-im
7096 Perform loop invariant motion on trees. This pass moves only invariants that
7097 would be hard to handle at RTL level (function calls, operations that expand to
7098 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7099 operands of conditions that are invariant out of the loop, so that we can use
7100 just trivial invariantness analysis in loop unswitching. The pass also includes
7103 @item -ftree-loop-ivcanon
7104 @opindex ftree-loop-ivcanon
7105 Create a canonical counter for number of iterations in the loop for that
7106 determining number of iterations requires complicated analysis. Later
7107 optimizations then may determine the number easily. Useful especially
7108 in connection with unrolling.
7112 Perform induction variable optimizations (strength reduction, induction
7113 variable merging and induction variable elimination) on trees.
7115 @item -ftree-parallelize-loops=n
7116 @opindex ftree-parallelize-loops
7117 Parallelize loops, i.e., split their iteration space to run in n threads.
7118 This is only possible for loops whose iterations are independent
7119 and can be arbitrarily reordered. The optimization is only
7120 profitable on multiprocessor machines, for loops that are CPU-intensive,
7121 rather than constrained e.g.@: by memory bandwidth. This option
7122 implies @option{-pthread}, and thus is only supported on targets
7123 that have support for @option{-pthread}.
7127 Perform function-local points-to analysis on trees. This flag is
7128 enabled by default at @option{-O} and higher.
7132 Perform scalar replacement of aggregates. This pass replaces structure
7133 references with scalars to prevent committing structures to memory too
7134 early. This flag is enabled by default at @option{-O} and higher.
7136 @item -ftree-copyrename
7137 @opindex ftree-copyrename
7138 Perform copy renaming on trees. This pass attempts to rename compiler
7139 temporaries to other variables at copy locations, usually resulting in
7140 variable names which more closely resemble the original variables. This flag
7141 is enabled by default at @option{-O} and higher.
7145 Perform temporary expression replacement during the SSA->normal phase. Single
7146 use/single def temporaries are replaced at their use location with their
7147 defining expression. This results in non-GIMPLE code, but gives the expanders
7148 much more complex trees to work on resulting in better RTL generation. This is
7149 enabled by default at @option{-O} and higher.
7151 @item -ftree-vectorize
7152 @opindex ftree-vectorize
7153 Perform loop vectorization on trees. This flag is enabled by default at
7156 @item -ftree-slp-vectorize
7157 @opindex ftree-slp-vectorize
7158 Perform basic block vectorization on trees. This flag is enabled by default at
7159 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7161 @item -ftree-vect-loop-version
7162 @opindex ftree-vect-loop-version
7163 Perform loop versioning when doing loop vectorization on trees. When a loop
7164 appears to be vectorizable except that data alignment or data dependence cannot
7165 be determined at compile time then vectorized and non-vectorized versions of
7166 the loop are generated along with runtime checks for alignment or dependence
7167 to control which version is executed. This option is enabled by default
7168 except at level @option{-Os} where it is disabled.
7170 @item -fvect-cost-model
7171 @opindex fvect-cost-model
7172 Enable cost model for vectorization.
7176 Perform Value Range Propagation on trees. This is similar to the
7177 constant propagation pass, but instead of values, ranges of values are
7178 propagated. This allows the optimizers to remove unnecessary range
7179 checks like array bound checks and null pointer checks. This is
7180 enabled by default at @option{-O2} and higher. Null pointer check
7181 elimination is only done if @option{-fdelete-null-pointer-checks} is
7186 Perform tail duplication to enlarge superblock size. This transformation
7187 simplifies the control flow of the function allowing other optimizations to do
7190 @item -funroll-loops
7191 @opindex funroll-loops
7192 Unroll loops whose number of iterations can be determined at compile
7193 time or upon entry to the loop. @option{-funroll-loops} implies
7194 @option{-frerun-cse-after-loop}. This option makes code larger,
7195 and may or may not make it run faster.
7197 @item -funroll-all-loops
7198 @opindex funroll-all-loops
7199 Unroll all loops, even if their number of iterations is uncertain when
7200 the loop is entered. This usually makes programs run more slowly.
7201 @option{-funroll-all-loops} implies the same options as
7202 @option{-funroll-loops},
7204 @item -fsplit-ivs-in-unroller
7205 @opindex fsplit-ivs-in-unroller
7206 Enables expressing of values of induction variables in later iterations
7207 of the unrolled loop using the value in the first iteration. This breaks
7208 long dependency chains, thus improving efficiency of the scheduling passes.
7210 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7211 same effect. However in cases the loop body is more complicated than
7212 a single basic block, this is not reliable. It also does not work at all
7213 on some of the architectures due to restrictions in the CSE pass.
7215 This optimization is enabled by default.
7217 @item -fvariable-expansion-in-unroller
7218 @opindex fvariable-expansion-in-unroller
7219 With this option, the compiler will create multiple copies of some
7220 local variables when unrolling a loop which can result in superior code.
7222 @item -fpartial-inlining
7223 @opindex fpartial-inlining
7224 Inline parts of functions. This option has any effect only
7225 when inlining itself is turned on by the @option{-finline-functions}
7226 or @option{-finline-small-functions} options.
7228 Enabled at level @option{-O2}.
7230 @item -fpredictive-commoning
7231 @opindex fpredictive-commoning
7232 Perform predictive commoning optimization, i.e., reusing computations
7233 (especially memory loads and stores) performed in previous
7234 iterations of loops.
7236 This option is enabled at level @option{-O3}.
7238 @item -fprefetch-loop-arrays
7239 @opindex fprefetch-loop-arrays
7240 If supported by the target machine, generate instructions to prefetch
7241 memory to improve the performance of loops that access large arrays.
7243 This option may generate better or worse code; results are highly
7244 dependent on the structure of loops within the source code.
7246 Disabled at level @option{-Os}.
7249 @itemx -fno-peephole2
7250 @opindex fno-peephole
7251 @opindex fno-peephole2
7252 Disable any machine-specific peephole optimizations. The difference
7253 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7254 are implemented in the compiler; some targets use one, some use the
7255 other, a few use both.
7257 @option{-fpeephole} is enabled by default.
7258 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7260 @item -fno-guess-branch-probability
7261 @opindex fno-guess-branch-probability
7262 Do not guess branch probabilities using heuristics.
7264 GCC will use heuristics to guess branch probabilities if they are
7265 not provided by profiling feedback (@option{-fprofile-arcs}). These
7266 heuristics are based on the control flow graph. If some branch probabilities
7267 are specified by @samp{__builtin_expect}, then the heuristics will be
7268 used to guess branch probabilities for the rest of the control flow graph,
7269 taking the @samp{__builtin_expect} info into account. The interactions
7270 between the heuristics and @samp{__builtin_expect} can be complex, and in
7271 some cases, it may be useful to disable the heuristics so that the effects
7272 of @samp{__builtin_expect} are easier to understand.
7274 The default is @option{-fguess-branch-probability} at levels
7275 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7277 @item -freorder-blocks
7278 @opindex freorder-blocks
7279 Reorder basic blocks in the compiled function in order to reduce number of
7280 taken branches and improve code locality.
7282 Enabled at levels @option{-O2}, @option{-O3}.
7284 @item -freorder-blocks-and-partition
7285 @opindex freorder-blocks-and-partition
7286 In addition to reordering basic blocks in the compiled function, in order
7287 to reduce number of taken branches, partitions hot and cold basic blocks
7288 into separate sections of the assembly and .o files, to improve
7289 paging and cache locality performance.
7291 This optimization is automatically turned off in the presence of
7292 exception handling, for linkonce sections, for functions with a user-defined
7293 section attribute and on any architecture that does not support named
7296 @item -freorder-functions
7297 @opindex freorder-functions
7298 Reorder functions in the object file in order to
7299 improve code locality. This is implemented by using special
7300 subsections @code{.text.hot} for most frequently executed functions and
7301 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7302 the linker so object file format must support named sections and linker must
7303 place them in a reasonable way.
7305 Also profile feedback must be available in to make this option effective. See
7306 @option{-fprofile-arcs} for details.
7308 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7310 @item -fstrict-aliasing
7311 @opindex fstrict-aliasing
7312 Allow the compiler to assume the strictest aliasing rules applicable to
7313 the language being compiled. For C (and C++), this activates
7314 optimizations based on the type of expressions. In particular, an
7315 object of one type is assumed never to reside at the same address as an
7316 object of a different type, unless the types are almost the same. For
7317 example, an @code{unsigned int} can alias an @code{int}, but not a
7318 @code{void*} or a @code{double}. A character type may alias any other
7321 @anchor{Type-punning}Pay special attention to code like this:
7334 The practice of reading from a different union member than the one most
7335 recently written to (called ``type-punning'') is common. Even with
7336 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7337 is accessed through the union type. So, the code above will work as
7338 expected. @xref{Structures unions enumerations and bit-fields
7339 implementation}. However, this code might not:
7350 Similarly, access by taking the address, casting the resulting pointer
7351 and dereferencing the result has undefined behavior, even if the cast
7352 uses a union type, e.g.:
7356 return ((union a_union *) &d)->i;
7360 The @option{-fstrict-aliasing} option is enabled at levels
7361 @option{-O2}, @option{-O3}, @option{-Os}.
7363 @item -fstrict-overflow
7364 @opindex fstrict-overflow
7365 Allow the compiler to assume strict signed overflow rules, depending
7366 on the language being compiled. For C (and C++) this means that
7367 overflow when doing arithmetic with signed numbers is undefined, which
7368 means that the compiler may assume that it will not happen. This
7369 permits various optimizations. For example, the compiler will assume
7370 that an expression like @code{i + 10 > i} will always be true for
7371 signed @code{i}. This assumption is only valid if signed overflow is
7372 undefined, as the expression is false if @code{i + 10} overflows when
7373 using twos complement arithmetic. When this option is in effect any
7374 attempt to determine whether an operation on signed numbers will
7375 overflow must be written carefully to not actually involve overflow.
7377 This option also allows the compiler to assume strict pointer
7378 semantics: given a pointer to an object, if adding an offset to that
7379 pointer does not produce a pointer to the same object, the addition is
7380 undefined. This permits the compiler to conclude that @code{p + u >
7381 p} is always true for a pointer @code{p} and unsigned integer
7382 @code{u}. This assumption is only valid because pointer wraparound is
7383 undefined, as the expression is false if @code{p + u} overflows using
7384 twos complement arithmetic.
7386 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7387 that integer signed overflow is fully defined: it wraps. When
7388 @option{-fwrapv} is used, there is no difference between
7389 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7390 integers. With @option{-fwrapv} certain types of overflow are
7391 permitted. For example, if the compiler gets an overflow when doing
7392 arithmetic on constants, the overflowed value can still be used with
7393 @option{-fwrapv}, but not otherwise.
7395 The @option{-fstrict-overflow} option is enabled at levels
7396 @option{-O2}, @option{-O3}, @option{-Os}.
7398 @item -falign-functions
7399 @itemx -falign-functions=@var{n}
7400 @opindex falign-functions
7401 Align the start of functions to the next power-of-two greater than
7402 @var{n}, skipping up to @var{n} bytes. For instance,
7403 @option{-falign-functions=32} aligns functions to the next 32-byte
7404 boundary, but @option{-falign-functions=24} would align to the next
7405 32-byte boundary only if this can be done by skipping 23 bytes or less.
7407 @option{-fno-align-functions} and @option{-falign-functions=1} are
7408 equivalent and mean that functions will not be aligned.
7410 Some assemblers only support this flag when @var{n} is a power of two;
7411 in that case, it is rounded up.
7413 If @var{n} is not specified or is zero, use a machine-dependent default.
7415 Enabled at levels @option{-O2}, @option{-O3}.
7417 @item -falign-labels
7418 @itemx -falign-labels=@var{n}
7419 @opindex falign-labels
7420 Align all branch targets to a power-of-two boundary, skipping up to
7421 @var{n} bytes like @option{-falign-functions}. This option can easily
7422 make code slower, because it must insert dummy operations for when the
7423 branch target is reached in the usual flow of the code.
7425 @option{-fno-align-labels} and @option{-falign-labels=1} are
7426 equivalent and mean that labels will not be aligned.
7428 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7429 are greater than this value, then their values are used instead.
7431 If @var{n} is not specified or is zero, use a machine-dependent default
7432 which is very likely to be @samp{1}, meaning no alignment.
7434 Enabled at levels @option{-O2}, @option{-O3}.
7437 @itemx -falign-loops=@var{n}
7438 @opindex falign-loops
7439 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7440 like @option{-falign-functions}. The hope is that the loop will be
7441 executed many times, which will make up for any execution of the dummy
7444 @option{-fno-align-loops} and @option{-falign-loops=1} are
7445 equivalent and mean that loops will not be aligned.
7447 If @var{n} is not specified or is zero, use a machine-dependent default.
7449 Enabled at levels @option{-O2}, @option{-O3}.
7452 @itemx -falign-jumps=@var{n}
7453 @opindex falign-jumps
7454 Align branch targets to a power-of-two boundary, for branch targets
7455 where the targets can only be reached by jumping, skipping up to @var{n}
7456 bytes like @option{-falign-functions}. In this case, no dummy operations
7459 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7460 equivalent and mean that loops will not be aligned.
7462 If @var{n} is not specified or is zero, use a machine-dependent default.
7464 Enabled at levels @option{-O2}, @option{-O3}.
7466 @item -funit-at-a-time
7467 @opindex funit-at-a-time
7468 This option is left for compatibility reasons. @option{-funit-at-a-time}
7469 has no effect, while @option{-fno-unit-at-a-time} implies
7470 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7474 @item -fno-toplevel-reorder
7475 @opindex fno-toplevel-reorder
7476 Do not reorder top-level functions, variables, and @code{asm}
7477 statements. Output them in the same order that they appear in the
7478 input file. When this option is used, unreferenced static variables
7479 will not be removed. This option is intended to support existing code
7480 which relies on a particular ordering. For new code, it is better to
7483 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7484 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7489 Constructs webs as commonly used for register allocation purposes and assign
7490 each web individual pseudo register. This allows the register allocation pass
7491 to operate on pseudos directly, but also strengthens several other optimization
7492 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7493 however, make debugging impossible, since variables will no longer stay in a
7496 Enabled by default with @option{-funroll-loops}.
7498 @item -fwhole-program
7499 @opindex fwhole-program
7500 Assume that the current compilation unit represents the whole program being
7501 compiled. All public functions and variables with the exception of @code{main}
7502 and those merged by attribute @code{externally_visible} become static functions
7503 and in effect are optimized more aggressively by interprocedural optimizers. If @command{gold} is used as the linker plugin, @code{externally_visible} attributes are automatically added to functions (not variable yet due to a current @command{gold} issue) that are accessed outside of LTO objects according to resolution file produced by @command{gold}. For other linkers that cannot generate resolution file, explicit @code{externally_visible} attributes are still necessary.
7504 While this option is equivalent to proper use of the @code{static} keyword for
7505 programs consisting of a single file, in combination with option
7506 @option{-flto} this flag can be used to
7507 compile many smaller scale programs since the functions and variables become
7508 local for the whole combined compilation unit, not for the single source file
7511 This option implies @option{-fwhole-file} for Fortran programs.
7513 @item -flto[=@var{n}]
7515 This option runs the standard link-time optimizer. When invoked
7516 with source code, it generates GIMPLE (one of GCC's internal
7517 representations) and writes it to special ELF sections in the object
7518 file. When the object files are linked together, all the function
7519 bodies are read from these ELF sections and instantiated as if they
7520 had been part of the same translation unit.
7522 To use the link-timer optimizer, @option{-flto} needs to be specified at
7523 compile time and during the final link. For example,
7526 gcc -c -O2 -flto foo.c
7527 gcc -c -O2 -flto bar.c
7528 gcc -o myprog -flto -O2 foo.o bar.o
7531 The first two invocations to GCC will save a bytecode representation
7532 of GIMPLE into special ELF sections inside @file{foo.o} and
7533 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7534 @file{foo.o} and @file{bar.o}, merge the two files into a single
7535 internal image, and compile the result as usual. Since both
7536 @file{foo.o} and @file{bar.o} are merged into a single image, this
7537 causes all the inter-procedural analyses and optimizations in GCC to
7538 work across the two files as if they were a single one. This means,
7539 for example, that the inliner will be able to inline functions in
7540 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7542 Another (simpler) way to enable link-time optimization is,
7545 gcc -o myprog -flto -O2 foo.c bar.c
7548 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7549 merge them together into a single GIMPLE representation and optimize
7550 them as usual to produce @file{myprog}.
7552 The only important thing to keep in mind is that to enable link-time
7553 optimizations the @option{-flto} flag needs to be passed to both the
7554 compile and the link commands.
7556 To make whole program optimization effective, it is necessary to make
7557 certain whole program assumptions. The compiler needs to know
7558 what functions and variables can be accessed by libraries and runtime
7559 outside of the link time optimized unit. When supported by the linker,
7560 the linker plugin (see @option{-fuse-linker-plugin}) passes to the
7561 compiler information about used and externally visible symbols. When
7562 the linker plugin is not available, @option{-fwhole-program} should be
7563 used to allow the compiler to make these assumptions, which will lead
7564 to more aggressive optimization decisions.
7566 Note that when a file is compiled with @option{-flto}, the generated
7567 object file will be larger than a regular object file because it will
7568 contain GIMPLE bytecodes and the usual final code. This means that
7569 object files with LTO information can be linked as a normal object
7570 file. So, in the previous example, if the final link is done with
7573 gcc -o myprog foo.o bar.o
7576 The only difference will be that no inter-procedural optimizations
7577 will be applied to produce @file{myprog}. The two object files
7578 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7581 Additionally, the optimization flags used to compile individual files
7582 are not necessarily related to those used at link-time. For instance,
7585 gcc -c -O0 -flto foo.c
7586 gcc -c -O0 -flto bar.c
7587 gcc -o myprog -flto -O3 foo.o bar.o
7590 This will produce individual object files with unoptimized assembler
7591 code, but the resulting binary @file{myprog} will be optimized at
7592 @option{-O3}. Now, if the final binary is generated without
7593 @option{-flto}, then @file{myprog} will not be optimized.
7595 When producing the final binary with @option{-flto}, GCC will only
7596 apply link-time optimizations to those files that contain bytecode.
7597 Therefore, you can mix and match object files and libraries with
7598 GIMPLE bytecodes and final object code. GCC will automatically select
7599 which files to optimize in LTO mode and which files to link without
7602 There are some code generation flags that GCC will preserve when
7603 generating bytecodes, as they need to be used during the final link
7604 stage. Currently, the following options are saved into the GIMPLE
7605 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7606 @option{-m} target flags.
7608 At link time, these options are read-in and reapplied. Note that the
7609 current implementation makes no attempt at recognizing conflicting
7610 values for these options. If two or more files have a conflicting
7611 value (e.g., one file is compiled with @option{-fPIC} and another
7612 isn't), the compiler will simply use the last value read from the
7613 bytecode files. It is recommended, then, that all the files
7614 participating in the same link be compiled with the same options.
7616 Another feature of LTO is that it is possible to apply interprocedural
7617 optimizations on files written in different languages. This requires
7618 some support in the language front end. Currently, the C, C++ and
7619 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7620 something like this should work
7625 gfortran -c -flto baz.f90
7626 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7629 Notice that the final link is done with @command{g++} to get the C++
7630 runtime libraries and @option{-lgfortran} is added to get the Fortran
7631 runtime libraries. In general, when mixing languages in LTO mode, you
7632 should use the same link command used when mixing languages in a
7633 regular (non-LTO) compilation. This means that if your build process
7634 was mixing languages before, all you need to add is @option{-flto} to
7635 all the compile and link commands.
7637 If LTO encounters objects with C linkage declared with incompatible
7638 types in separate translation units to be linked together (undefined
7639 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7640 issued. The behavior is still undefined at runtime.
7642 If object files containing GIMPLE bytecode are stored in a library archive, say
7643 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
7644 are using a linker with linker plugin support. To enable this feature, use
7645 the flag @option{-fuse-linker-plugin} at link-time:
7648 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7651 With the linker plugin enabled, the linker will extract the needed
7652 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7653 to make them part of the aggregated GIMPLE image to be optimized.
7655 If you are not using a linker with linker plugin support and/or do not
7656 enable linker plugin then the objects inside @file{libfoo.a}
7657 will be extracted and linked as usual, but they will not participate
7658 in the LTO optimization process.
7660 Link time optimizations do not require the presence of the whole program to
7661 operate. If the program does not require any symbols to be exported, it is
7662 possible to combine @option{-flto} and with @option{-fwhole-program} to allow
7663 the interprocedural optimizers to use more aggressive assumptions which may
7664 lead to improved optimization opportunities.
7665 Use of @option{-fwhole-program} is not needed when linker plugin is
7666 active (see @option{-fuse-linker-plugin}).
7668 Regarding portability: the current implementation of LTO makes no
7669 attempt at generating bytecode that can be ported between different
7670 types of hosts. The bytecode files are versioned and there is a
7671 strict version check, so bytecode files generated in one version of
7672 GCC will not work with an older/newer version of GCC.
7674 Link time optimization does not play well with generating debugging
7675 information. Combining @option{-flto} with
7676 @option{-g} is currently experimental and expected to produce wrong
7679 If you specify the optional @var{n}, the optimization and code
7680 generation done at link time is executed in parallel using @var{n}
7681 parallel jobs by utilizing an installed @command{make} program. The
7682 environment variable @env{MAKE} may be used to override the program
7683 used. The default value for @var{n} is 1.
7685 You can also specify @option{-flto=jobserver} to use GNU make's
7686 job server mode to determine the number of parallel jobs. This
7687 is useful when the Makefile calling GCC is already executing in parallel.
7688 The parent Makefile will need a @samp{+} prepended to the command recipe
7689 for this to work. This will likely only work if @env{MAKE} is
7692 This option is disabled by default.
7694 @item -flto-partition=@var{alg}
7695 @opindex flto-partition
7696 Specify the partitioning algorithm used by the link time optimizer.
7697 The value is either @code{1to1} to specify a partitioning mirroring
7698 the original source files or @code{balanced} to specify partitioning
7699 into equally sized chunks (whenever possible). Specifying @code{none}
7700 as an algorithm disables partitioning and streaming completely. The
7701 default value is @code{balanced}.
7703 @item -flto-compression-level=@var{n}
7704 This option specifies the level of compression used for intermediate
7705 language written to LTO object files, and is only meaningful in
7706 conjunction with LTO mode (@option{-flto}). Valid
7707 values are 0 (no compression) to 9 (maximum compression). Values
7708 outside this range are clamped to either 0 or 9. If the option is not
7709 given, a default balanced compression setting is used.
7712 Prints a report with internal details on the workings of the link-time
7713 optimizer. The contents of this report vary from version to version,
7714 it is meant to be useful to GCC developers when processing object
7715 files in LTO mode (via @option{-flto}).
7717 Disabled by default.
7719 @item -fuse-linker-plugin
7720 Enables the use of linker plugin during link time optimization. This option
7721 relies on the linker plugin support in linker that is available in @code{gold}
7722 or in GNU ld 2.21.51 or newer..
7724 This option enables the extraction of object files with GIMPLE bytecode out of
7725 library archives. This improves the quality of optimization by exposing more
7726 code the the link time optimizer. This information specify what symbols
7727 can be accessed externally (by non-LTO object or during dynamic linking).
7728 Resulting code quality improvements on binaries (and shared libraries that do
7729 use hidden visibility) is similar to @code{-fwhole-program}. See
7730 @option{-flto} for a description on the effect of this flag and how to use it.
7732 Enabled by default when LTO support in GCC is enabled and GCC was compiled
7733 with linker supporting plugins (GNU ld or @code{gold}).
7735 @item -fcompare-elim
7736 @opindex fcompare-elim
7737 After register allocation and post-register allocation instruction splitting,
7738 identify arithmetic instructions that compute processor flags similar to a
7739 comparison operation based on that arithmetic. If possible, eliminate the
7740 explicit comparison operation.
7742 This pass only applies to certain targets that cannot explicitly represent
7743 the comparison operation before register allocation is complete.
7745 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7747 @item -fcprop-registers
7748 @opindex fcprop-registers
7749 After register allocation and post-register allocation instruction splitting,
7750 we perform a copy-propagation pass to try to reduce scheduling dependencies
7751 and occasionally eliminate the copy.
7753 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7755 @item -fprofile-correction
7756 @opindex fprofile-correction
7757 Profiles collected using an instrumented binary for multi-threaded programs may
7758 be inconsistent due to missed counter updates. When this option is specified,
7759 GCC will use heuristics to correct or smooth out such inconsistencies. By
7760 default, GCC will emit an error message when an inconsistent profile is detected.
7762 @item -fprofile-dir=@var{path}
7763 @opindex fprofile-dir
7765 Set the directory to search for the profile data files in to @var{path}.
7766 This option affects only the profile data generated by
7767 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7768 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7769 and its related options.
7770 By default, GCC will use the current directory as @var{path}, thus the
7771 profile data file will appear in the same directory as the object file.
7773 @item -fprofile-generate
7774 @itemx -fprofile-generate=@var{path}
7775 @opindex fprofile-generate
7777 Enable options usually used for instrumenting application to produce
7778 profile useful for later recompilation with profile feedback based
7779 optimization. You must use @option{-fprofile-generate} both when
7780 compiling and when linking your program.
7782 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7784 If @var{path} is specified, GCC will look at the @var{path} to find
7785 the profile feedback data files. See @option{-fprofile-dir}.
7788 @itemx -fprofile-use=@var{path}
7789 @opindex fprofile-use
7790 Enable profile feedback directed optimizations, and optimizations
7791 generally profitable only with profile feedback available.
7793 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7794 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7796 By default, GCC emits an error message if the feedback profiles do not
7797 match the source code. This error can be turned into a warning by using
7798 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7801 If @var{path} is specified, GCC will look at the @var{path} to find
7802 the profile feedback data files. See @option{-fprofile-dir}.
7805 The following options control compiler behavior regarding floating
7806 point arithmetic. These options trade off between speed and
7807 correctness. All must be specifically enabled.
7811 @opindex ffloat-store
7812 Do not store floating point variables in registers, and inhibit other
7813 options that might change whether a floating point value is taken from a
7816 @cindex floating point precision
7817 This option prevents undesirable excess precision on machines such as
7818 the 68000 where the floating registers (of the 68881) keep more
7819 precision than a @code{double} is supposed to have. Similarly for the
7820 x86 architecture. For most programs, the excess precision does only
7821 good, but a few programs rely on the precise definition of IEEE floating
7822 point. Use @option{-ffloat-store} for such programs, after modifying
7823 them to store all pertinent intermediate computations into variables.
7825 @item -fexcess-precision=@var{style}
7826 @opindex fexcess-precision
7827 This option allows further control over excess precision on machines
7828 where floating-point registers have more precision than the IEEE
7829 @code{float} and @code{double} types and the processor does not
7830 support operations rounding to those types. By default,
7831 @option{-fexcess-precision=fast} is in effect; this means that
7832 operations are carried out in the precision of the registers and that
7833 it is unpredictable when rounding to the types specified in the source
7834 code takes place. When compiling C, if
7835 @option{-fexcess-precision=standard} is specified then excess
7836 precision will follow the rules specified in ISO C99; in particular,
7837 both casts and assignments cause values to be rounded to their
7838 semantic types (whereas @option{-ffloat-store} only affects
7839 assignments). This option is enabled by default for C if a strict
7840 conformance option such as @option{-std=c99} is used.
7843 @option{-fexcess-precision=standard} is not implemented for languages
7844 other than C, and has no effect if
7845 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7846 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7847 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7848 semantics apply without excess precision, and in the latter, rounding
7853 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7854 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7855 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7857 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7859 This option is not turned on by any @option{-O} option besides
7860 @option{-Ofast} since it can result in incorrect output for programs
7861 which depend on an exact implementation of IEEE or ISO rules/specifications
7862 for math functions. It may, however, yield faster code for programs
7863 that do not require the guarantees of these specifications.
7865 @item -fno-math-errno
7866 @opindex fno-math-errno
7867 Do not set ERRNO after calling math functions that are executed
7868 with a single instruction, e.g., sqrt. A program that relies on
7869 IEEE exceptions for math error handling may want to use this flag
7870 for speed while maintaining IEEE arithmetic compatibility.
7872 This option is not turned on by any @option{-O} option since
7873 it can result in incorrect output for programs which depend on
7874 an exact implementation of IEEE or ISO rules/specifications for
7875 math functions. It may, however, yield faster code for programs
7876 that do not require the guarantees of these specifications.
7878 The default is @option{-fmath-errno}.
7880 On Darwin systems, the math library never sets @code{errno}. There is
7881 therefore no reason for the compiler to consider the possibility that
7882 it might, and @option{-fno-math-errno} is the default.
7884 @item -funsafe-math-optimizations
7885 @opindex funsafe-math-optimizations
7887 Allow optimizations for floating-point arithmetic that (a) assume
7888 that arguments and results are valid and (b) may violate IEEE or
7889 ANSI standards. When used at link-time, it may include libraries
7890 or startup files that change the default FPU control word or other
7891 similar optimizations.
7893 This option is not turned on by any @option{-O} option since
7894 it can result in incorrect output for programs which depend on
7895 an exact implementation of IEEE or ISO rules/specifications for
7896 math functions. It may, however, yield faster code for programs
7897 that do not require the guarantees of these specifications.
7898 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7899 @option{-fassociative-math} and @option{-freciprocal-math}.
7901 The default is @option{-fno-unsafe-math-optimizations}.
7903 @item -fassociative-math
7904 @opindex fassociative-math
7906 Allow re-association of operands in series of floating-point operations.
7907 This violates the ISO C and C++ language standard by possibly changing
7908 computation result. NOTE: re-ordering may change the sign of zero as
7909 well as ignore NaNs and inhibit or create underflow or overflow (and
7910 thus cannot be used on a code which relies on rounding behavior like
7911 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7912 and thus may not be used when ordered comparisons are required.
7913 This option requires that both @option{-fno-signed-zeros} and
7914 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7915 much sense with @option{-frounding-math}. For Fortran the option
7916 is automatically enabled when both @option{-fno-signed-zeros} and
7917 @option{-fno-trapping-math} are in effect.
7919 The default is @option{-fno-associative-math}.
7921 @item -freciprocal-math
7922 @opindex freciprocal-math
7924 Allow the reciprocal of a value to be used instead of dividing by
7925 the value if this enables optimizations. For example @code{x / y}
7926 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7927 is subject to common subexpression elimination. Note that this loses
7928 precision and increases the number of flops operating on the value.
7930 The default is @option{-fno-reciprocal-math}.
7932 @item -ffinite-math-only
7933 @opindex ffinite-math-only
7934 Allow optimizations for floating-point arithmetic that assume
7935 that arguments and results are not NaNs or +-Infs.
7937 This option is not turned on by any @option{-O} option since
7938 it can result in incorrect output for programs which depend on
7939 an exact implementation of IEEE or ISO rules/specifications for
7940 math functions. It may, however, yield faster code for programs
7941 that do not require the guarantees of these specifications.
7943 The default is @option{-fno-finite-math-only}.
7945 @item -fno-signed-zeros
7946 @opindex fno-signed-zeros
7947 Allow optimizations for floating point arithmetic that ignore the
7948 signedness of zero. IEEE arithmetic specifies the behavior of
7949 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7950 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7951 This option implies that the sign of a zero result isn't significant.
7953 The default is @option{-fsigned-zeros}.
7955 @item -fno-trapping-math
7956 @opindex fno-trapping-math
7957 Compile code assuming that floating-point operations cannot generate
7958 user-visible traps. These traps include division by zero, overflow,
7959 underflow, inexact result and invalid operation. This option requires
7960 that @option{-fno-signaling-nans} be in effect. Setting this option may
7961 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7963 This option should never be turned on by any @option{-O} option since
7964 it can result in incorrect output for programs which depend on
7965 an exact implementation of IEEE or ISO rules/specifications for
7968 The default is @option{-ftrapping-math}.
7970 @item -frounding-math
7971 @opindex frounding-math
7972 Disable transformations and optimizations that assume default floating
7973 point rounding behavior. This is round-to-zero for all floating point
7974 to integer conversions, and round-to-nearest for all other arithmetic
7975 truncations. This option should be specified for programs that change
7976 the FP rounding mode dynamically, or that may be executed with a
7977 non-default rounding mode. This option disables constant folding of
7978 floating point expressions at compile-time (which may be affected by
7979 rounding mode) and arithmetic transformations that are unsafe in the
7980 presence of sign-dependent rounding modes.
7982 The default is @option{-fno-rounding-math}.
7984 This option is experimental and does not currently guarantee to
7985 disable all GCC optimizations that are affected by rounding mode.
7986 Future versions of GCC may provide finer control of this setting
7987 using C99's @code{FENV_ACCESS} pragma. This command line option
7988 will be used to specify the default state for @code{FENV_ACCESS}.
7990 @item -fsignaling-nans
7991 @opindex fsignaling-nans
7992 Compile code assuming that IEEE signaling NaNs may generate user-visible
7993 traps during floating-point operations. Setting this option disables
7994 optimizations that may change the number of exceptions visible with
7995 signaling NaNs. This option implies @option{-ftrapping-math}.
7997 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
8000 The default is @option{-fno-signaling-nans}.
8002 This option is experimental and does not currently guarantee to
8003 disable all GCC optimizations that affect signaling NaN behavior.
8005 @item -fsingle-precision-constant
8006 @opindex fsingle-precision-constant
8007 Treat floating point constant as single precision constant instead of
8008 implicitly converting it to double precision constant.
8010 @item -fcx-limited-range
8011 @opindex fcx-limited-range
8012 When enabled, this option states that a range reduction step is not
8013 needed when performing complex division. Also, there is no checking
8014 whether the result of a complex multiplication or division is @code{NaN
8015 + I*NaN}, with an attempt to rescue the situation in that case. The
8016 default is @option{-fno-cx-limited-range}, but is enabled by
8017 @option{-ffast-math}.
8019 This option controls the default setting of the ISO C99
8020 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8023 @item -fcx-fortran-rules
8024 @opindex fcx-fortran-rules
8025 Complex multiplication and division follow Fortran rules. Range
8026 reduction is done as part of complex division, but there is no checking
8027 whether the result of a complex multiplication or division is @code{NaN
8028 + I*NaN}, with an attempt to rescue the situation in that case.
8030 The default is @option{-fno-cx-fortran-rules}.
8034 The following options control optimizations that may improve
8035 performance, but are not enabled by any @option{-O} options. This
8036 section includes experimental options that may produce broken code.
8039 @item -fbranch-probabilities
8040 @opindex fbranch-probabilities
8041 After running a program compiled with @option{-fprofile-arcs}
8042 (@pxref{Debugging Options,, Options for Debugging Your Program or
8043 @command{gcc}}), you can compile it a second time using
8044 @option{-fbranch-probabilities}, to improve optimizations based on
8045 the number of times each branch was taken. When the program
8046 compiled with @option{-fprofile-arcs} exits it saves arc execution
8047 counts to a file called @file{@var{sourcename}.gcda} for each source
8048 file. The information in this data file is very dependent on the
8049 structure of the generated code, so you must use the same source code
8050 and the same optimization options for both compilations.
8052 With @option{-fbranch-probabilities}, GCC puts a
8053 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8054 These can be used to improve optimization. Currently, they are only
8055 used in one place: in @file{reorg.c}, instead of guessing which path a
8056 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8057 exactly determine which path is taken more often.
8059 @item -fprofile-values
8060 @opindex fprofile-values
8061 If combined with @option{-fprofile-arcs}, it adds code so that some
8062 data about values of expressions in the program is gathered.
8064 With @option{-fbranch-probabilities}, it reads back the data gathered
8065 from profiling values of expressions for usage in optimizations.
8067 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8071 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8072 a code to gather information about values of expressions.
8074 With @option{-fbranch-probabilities}, it reads back the data gathered
8075 and actually performs the optimizations based on them.
8076 Currently the optimizations include specialization of division operation
8077 using the knowledge about the value of the denominator.
8079 @item -frename-registers
8080 @opindex frename-registers
8081 Attempt to avoid false dependencies in scheduled code by making use
8082 of registers left over after register allocation. This optimization
8083 will most benefit processors with lots of registers. Depending on the
8084 debug information format adopted by the target, however, it can
8085 make debugging impossible, since variables will no longer stay in
8086 a ``home register''.
8088 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8092 Perform tail duplication to enlarge superblock size. This transformation
8093 simplifies the control flow of the function allowing other optimizations to do
8096 Enabled with @option{-fprofile-use}.
8098 @item -funroll-loops
8099 @opindex funroll-loops
8100 Unroll loops whose number of iterations can be determined at compile time or
8101 upon entry to the loop. @option{-funroll-loops} implies
8102 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8103 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8104 small constant number of iterations). This option makes code larger, and may
8105 or may not make it run faster.
8107 Enabled with @option{-fprofile-use}.
8109 @item -funroll-all-loops
8110 @opindex funroll-all-loops
8111 Unroll all loops, even if their number of iterations is uncertain when
8112 the loop is entered. This usually makes programs run more slowly.
8113 @option{-funroll-all-loops} implies the same options as
8114 @option{-funroll-loops}.
8117 @opindex fpeel-loops
8118 Peels the loops for that there is enough information that they do not
8119 roll much (from profile feedback). It also turns on complete loop peeling
8120 (i.e.@: complete removal of loops with small constant number of iterations).
8122 Enabled with @option{-fprofile-use}.
8124 @item -fmove-loop-invariants
8125 @opindex fmove-loop-invariants
8126 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8127 at level @option{-O1}
8129 @item -funswitch-loops
8130 @opindex funswitch-loops
8131 Move branches with loop invariant conditions out of the loop, with duplicates
8132 of the loop on both branches (modified according to result of the condition).
8134 @item -ffunction-sections
8135 @itemx -fdata-sections
8136 @opindex ffunction-sections
8137 @opindex fdata-sections
8138 Place each function or data item into its own section in the output
8139 file if the target supports arbitrary sections. The name of the
8140 function or the name of the data item determines the section's name
8143 Use these options on systems where the linker can perform optimizations
8144 to improve locality of reference in the instruction space. Most systems
8145 using the ELF object format and SPARC processors running Solaris 2 have
8146 linkers with such optimizations. AIX may have these optimizations in
8149 Only use these options when there are significant benefits from doing
8150 so. When you specify these options, the assembler and linker will
8151 create larger object and executable files and will also be slower.
8152 You will not be able to use @code{gprof} on all systems if you
8153 specify this option and you may have problems with debugging if
8154 you specify both this option and @option{-g}.
8156 @item -fbranch-target-load-optimize
8157 @opindex fbranch-target-load-optimize
8158 Perform branch target register load optimization before prologue / epilogue
8160 The use of target registers can typically be exposed only during reload,
8161 thus hoisting loads out of loops and doing inter-block scheduling needs
8162 a separate optimization pass.
8164 @item -fbranch-target-load-optimize2
8165 @opindex fbranch-target-load-optimize2
8166 Perform branch target register load optimization after prologue / epilogue
8169 @item -fbtr-bb-exclusive
8170 @opindex fbtr-bb-exclusive
8171 When performing branch target register load optimization, don't reuse
8172 branch target registers in within any basic block.
8174 @item -fstack-protector
8175 @opindex fstack-protector
8176 Emit extra code to check for buffer overflows, such as stack smashing
8177 attacks. This is done by adding a guard variable to functions with
8178 vulnerable objects. This includes functions that call alloca, and
8179 functions with buffers larger than 8 bytes. The guards are initialized
8180 when a function is entered and then checked when the function exits.
8181 If a guard check fails, an error message is printed and the program exits.
8183 @item -fstack-protector-all
8184 @opindex fstack-protector-all
8185 Like @option{-fstack-protector} except that all functions are protected.
8187 @item -fsection-anchors
8188 @opindex fsection-anchors
8189 Try to reduce the number of symbolic address calculations by using
8190 shared ``anchor'' symbols to address nearby objects. This transformation
8191 can help to reduce the number of GOT entries and GOT accesses on some
8194 For example, the implementation of the following function @code{foo}:
8198 int foo (void) @{ return a + b + c; @}
8201 would usually calculate the addresses of all three variables, but if you
8202 compile it with @option{-fsection-anchors}, it will access the variables
8203 from a common anchor point instead. The effect is similar to the
8204 following pseudocode (which isn't valid C):
8209 register int *xr = &x;
8210 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8214 Not all targets support this option.
8216 @item --param @var{name}=@var{value}
8218 In some places, GCC uses various constants to control the amount of
8219 optimization that is done. For example, GCC will not inline functions
8220 that contain more that a certain number of instructions. You can
8221 control some of these constants on the command-line using the
8222 @option{--param} option.
8224 The names of specific parameters, and the meaning of the values, are
8225 tied to the internals of the compiler, and are subject to change
8226 without notice in future releases.
8228 In each case, the @var{value} is an integer. The allowable choices for
8229 @var{name} are given in the following table:
8232 @item struct-reorg-cold-struct-ratio
8233 The threshold ratio (as a percentage) between a structure frequency
8234 and the frequency of the hottest structure in the program. This parameter
8235 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
8236 We say that if the ratio of a structure frequency, calculated by profiling,
8237 to the hottest structure frequency in the program is less than this
8238 parameter, then structure reorganization is not applied to this structure.
8241 @item predictable-branch-outcome
8242 When branch is predicted to be taken with probability lower than this threshold
8243 (in percent), then it is considered well predictable. The default is 10.
8245 @item max-crossjump-edges
8246 The maximum number of incoming edges to consider for crossjumping.
8247 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8248 the number of edges incoming to each block. Increasing values mean
8249 more aggressive optimization, making the compile time increase with
8250 probably small improvement in executable size.
8252 @item min-crossjump-insns
8253 The minimum number of instructions which must be matched at the end
8254 of two blocks before crossjumping will be performed on them. This
8255 value is ignored in the case where all instructions in the block being
8256 crossjumped from are matched. The default value is 5.
8258 @item max-grow-copy-bb-insns
8259 The maximum code size expansion factor when copying basic blocks
8260 instead of jumping. The expansion is relative to a jump instruction.
8261 The default value is 8.
8263 @item max-goto-duplication-insns
8264 The maximum number of instructions to duplicate to a block that jumps
8265 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8266 passes, GCC factors computed gotos early in the compilation process,
8267 and unfactors them as late as possible. Only computed jumps at the
8268 end of a basic blocks with no more than max-goto-duplication-insns are
8269 unfactored. The default value is 8.
8271 @item max-delay-slot-insn-search
8272 The maximum number of instructions to consider when looking for an
8273 instruction to fill a delay slot. If more than this arbitrary number of
8274 instructions is searched, the time savings from filling the delay slot
8275 will be minimal so stop searching. Increasing values mean more
8276 aggressive optimization, making the compile time increase with probably
8277 small improvement in executable run time.
8279 @item max-delay-slot-live-search
8280 When trying to fill delay slots, the maximum number of instructions to
8281 consider when searching for a block with valid live register
8282 information. Increasing this arbitrarily chosen value means more
8283 aggressive optimization, increasing the compile time. This parameter
8284 should be removed when the delay slot code is rewritten to maintain the
8287 @item max-gcse-memory
8288 The approximate maximum amount of memory that will be allocated in
8289 order to perform the global common subexpression elimination
8290 optimization. If more memory than specified is required, the
8291 optimization will not be done.
8293 @item max-gcse-insertion-ratio
8294 If the ratio of expression insertions to deletions is larger than this value
8295 for any expression, then RTL PRE will insert or remove the expression and thus
8296 leave partially redundant computations in the instruction stream. The default value is 20.
8298 @item max-pending-list-length
8299 The maximum number of pending dependencies scheduling will allow
8300 before flushing the current state and starting over. Large functions
8301 with few branches or calls can create excessively large lists which
8302 needlessly consume memory and resources.
8304 @item max-inline-insns-single
8305 Several parameters control the tree inliner used in gcc.
8306 This number sets the maximum number of instructions (counted in GCC's
8307 internal representation) in a single function that the tree inliner
8308 will consider for inlining. This only affects functions declared
8309 inline and methods implemented in a class declaration (C++).
8310 The default value is 400.
8312 @item max-inline-insns-auto
8313 When you use @option{-finline-functions} (included in @option{-O3}),
8314 a lot of functions that would otherwise not be considered for inlining
8315 by the compiler will be investigated. To those functions, a different
8316 (more restrictive) limit compared to functions declared inline can
8318 The default value is 40.
8320 @item large-function-insns
8321 The limit specifying really large functions. For functions larger than this
8322 limit after inlining, inlining is constrained by
8323 @option{--param large-function-growth}. This parameter is useful primarily
8324 to avoid extreme compilation time caused by non-linear algorithms used by the
8326 The default value is 2700.
8328 @item large-function-growth
8329 Specifies maximal growth of large function caused by inlining in percents.
8330 The default value is 100 which limits large function growth to 2.0 times
8333 @item large-unit-insns
8334 The limit specifying large translation unit. Growth caused by inlining of
8335 units larger than this limit is limited by @option{--param inline-unit-growth}.
8336 For small units this might be too tight (consider unit consisting of function A
8337 that is inline and B that just calls A three time. If B is small relative to
8338 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8339 large units consisting of small inlineable functions however the overall unit
8340 growth limit is needed to avoid exponential explosion of code size. Thus for
8341 smaller units, the size is increased to @option{--param large-unit-insns}
8342 before applying @option{--param inline-unit-growth}. The default is 10000
8344 @item inline-unit-growth
8345 Specifies maximal overall growth of the compilation unit caused by inlining.
8346 The default value is 30 which limits unit growth to 1.3 times the original
8349 @item ipcp-unit-growth
8350 Specifies maximal overall growth of the compilation unit caused by
8351 interprocedural constant propagation. The default value is 10 which limits
8352 unit growth to 1.1 times the original size.
8354 @item large-stack-frame
8355 The limit specifying large stack frames. While inlining the algorithm is trying
8356 to not grow past this limit too much. Default value is 256 bytes.
8358 @item large-stack-frame-growth
8359 Specifies maximal growth of large stack frames caused by inlining in percents.
8360 The default value is 1000 which limits large stack frame growth to 11 times
8363 @item max-inline-insns-recursive
8364 @itemx max-inline-insns-recursive-auto
8365 Specifies maximum number of instructions out-of-line copy of self recursive inline
8366 function can grow into by performing recursive inlining.
8368 For functions declared inline @option{--param max-inline-insns-recursive} is
8369 taken into account. For function not declared inline, recursive inlining
8370 happens only when @option{-finline-functions} (included in @option{-O3}) is
8371 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8372 default value is 450.
8374 @item max-inline-recursive-depth
8375 @itemx max-inline-recursive-depth-auto
8376 Specifies maximum recursion depth used by the recursive inlining.
8378 For functions declared inline @option{--param max-inline-recursive-depth} is
8379 taken into account. For function not declared inline, recursive inlining
8380 happens only when @option{-finline-functions} (included in @option{-O3}) is
8381 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8384 @item min-inline-recursive-probability
8385 Recursive inlining is profitable only for function having deep recursion
8386 in average and can hurt for function having little recursion depth by
8387 increasing the prologue size or complexity of function body to other
8390 When profile feedback is available (see @option{-fprofile-generate}) the actual
8391 recursion depth can be guessed from probability that function will recurse via
8392 given call expression. This parameter limits inlining only to call expression
8393 whose probability exceeds given threshold (in percents). The default value is
8396 @item early-inlining-insns
8397 Specify growth that early inliner can make. In effect it increases amount of
8398 inlining for code having large abstraction penalty. The default value is 10.
8400 @item max-early-inliner-iterations
8401 @itemx max-early-inliner-iterations
8402 Limit of iterations of early inliner. This basically bounds number of nested
8403 indirect calls early inliner can resolve. Deeper chains are still handled by
8406 @item comdat-sharing-probability
8407 @itemx comdat-sharing-probability
8408 Probability (in percent) that C++ inline function with comdat visibility
8409 will be shared across multiple compilation units. The default value is 20.
8411 @item min-vect-loop-bound
8412 The minimum number of iterations under which a loop will not get vectorized
8413 when @option{-ftree-vectorize} is used. The number of iterations after
8414 vectorization needs to be greater than the value specified by this option
8415 to allow vectorization. The default value is 0.
8417 @item gcse-cost-distance-ratio
8418 Scaling factor in calculation of maximum distance an expression
8419 can be moved by GCSE optimizations. This is currently supported only in the
8420 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8421 will be with simple expressions, i.e., the expressions which have cost
8422 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8423 hoisting of simple expressions. The default value is 10.
8425 @item gcse-unrestricted-cost
8426 Cost, roughly measured as the cost of a single typical machine
8427 instruction, at which GCSE optimizations will not constrain
8428 the distance an expression can travel. This is currently
8429 supported only in the code hoisting pass. The lesser the cost,
8430 the more aggressive code hoisting will be. Specifying 0 will
8431 allow all expressions to travel unrestricted distances.
8432 The default value is 3.
8434 @item max-hoist-depth
8435 The depth of search in the dominator tree for expressions to hoist.
8436 This is used to avoid quadratic behavior in hoisting algorithm.
8437 The value of 0 will avoid limiting the search, but may slow down compilation
8438 of huge functions. The default value is 30.
8440 @item max-unrolled-insns
8441 The maximum number of instructions that a loop should have if that loop
8442 is unrolled, and if the loop is unrolled, it determines how many times
8443 the loop code is unrolled.
8445 @item max-average-unrolled-insns
8446 The maximum number of instructions biased by probabilities of their execution
8447 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8448 it determines how many times the loop code is unrolled.
8450 @item max-unroll-times
8451 The maximum number of unrollings of a single loop.
8453 @item max-peeled-insns
8454 The maximum number of instructions that a loop should have if that loop
8455 is peeled, and if the loop is peeled, it determines how many times
8456 the loop code is peeled.
8458 @item max-peel-times
8459 The maximum number of peelings of a single loop.
8461 @item max-completely-peeled-insns
8462 The maximum number of insns of a completely peeled loop.
8464 @item max-completely-peel-times
8465 The maximum number of iterations of a loop to be suitable for complete peeling.
8467 @item max-completely-peel-loop-nest-depth
8468 The maximum depth of a loop nest suitable for complete peeling.
8470 @item max-unswitch-insns
8471 The maximum number of insns of an unswitched loop.
8473 @item max-unswitch-level
8474 The maximum number of branches unswitched in a single loop.
8477 The minimum cost of an expensive expression in the loop invariant motion.
8479 @item iv-consider-all-candidates-bound
8480 Bound on number of candidates for induction variables below that
8481 all candidates are considered for each use in induction variable
8482 optimizations. Only the most relevant candidates are considered
8483 if there are more candidates, to avoid quadratic time complexity.
8485 @item iv-max-considered-uses
8486 The induction variable optimizations give up on loops that contain more
8487 induction variable uses.
8489 @item iv-always-prune-cand-set-bound
8490 If number of candidates in the set is smaller than this value,
8491 we always try to remove unnecessary ivs from the set during its
8492 optimization when a new iv is added to the set.
8494 @item scev-max-expr-size
8495 Bound on size of expressions used in the scalar evolutions analyzer.
8496 Large expressions slow the analyzer.
8498 @item scev-max-expr-complexity
8499 Bound on the complexity of the expressions in the scalar evolutions analyzer.
8500 Complex expressions slow the analyzer.
8502 @item omega-max-vars
8503 The maximum number of variables in an Omega constraint system.
8504 The default value is 128.
8506 @item omega-max-geqs
8507 The maximum number of inequalities in an Omega constraint system.
8508 The default value is 256.
8511 The maximum number of equalities in an Omega constraint system.
8512 The default value is 128.
8514 @item omega-max-wild-cards
8515 The maximum number of wildcard variables that the Omega solver will
8516 be able to insert. The default value is 18.
8518 @item omega-hash-table-size
8519 The size of the hash table in the Omega solver. The default value is
8522 @item omega-max-keys
8523 The maximal number of keys used by the Omega solver. The default
8526 @item omega-eliminate-redundant-constraints
8527 When set to 1, use expensive methods to eliminate all redundant
8528 constraints. The default value is 0.
8530 @item vect-max-version-for-alignment-checks
8531 The maximum number of runtime checks that can be performed when
8532 doing loop versioning for alignment in the vectorizer. See option
8533 ftree-vect-loop-version for more information.
8535 @item vect-max-version-for-alias-checks
8536 The maximum number of runtime checks that can be performed when
8537 doing loop versioning for alias in the vectorizer. See option
8538 ftree-vect-loop-version for more information.
8540 @item max-iterations-to-track
8542 The maximum number of iterations of a loop the brute force algorithm
8543 for analysis of # of iterations of the loop tries to evaluate.
8545 @item hot-bb-count-fraction
8546 Select fraction of the maximal count of repetitions of basic block in program
8547 given basic block needs to have to be considered hot.
8549 @item hot-bb-frequency-fraction
8550 Select fraction of the entry block frequency of executions of basic block in
8551 function given basic block needs to have to be considered hot
8553 @item max-predicted-iterations
8554 The maximum number of loop iterations we predict statically. This is useful
8555 in cases where function contain single loop with known bound and other loop
8556 with unknown. We predict the known number of iterations correctly, while
8557 the unknown number of iterations average to roughly 10. This means that the
8558 loop without bounds would appear artificially cold relative to the other one.
8560 @item align-threshold
8562 Select fraction of the maximal frequency of executions of basic block in
8563 function given basic block will get aligned.
8565 @item align-loop-iterations
8567 A loop expected to iterate at lest the selected number of iterations will get
8570 @item tracer-dynamic-coverage
8571 @itemx tracer-dynamic-coverage-feedback
8573 This value is used to limit superblock formation once the given percentage of
8574 executed instructions is covered. This limits unnecessary code size
8577 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8578 feedback is available. The real profiles (as opposed to statically estimated
8579 ones) are much less balanced allowing the threshold to be larger value.
8581 @item tracer-max-code-growth
8582 Stop tail duplication once code growth has reached given percentage. This is
8583 rather hokey argument, as most of the duplicates will be eliminated later in
8584 cross jumping, so it may be set to much higher values than is the desired code
8587 @item tracer-min-branch-ratio
8589 Stop reverse growth when the reverse probability of best edge is less than this
8590 threshold (in percent).
8592 @item tracer-min-branch-ratio
8593 @itemx tracer-min-branch-ratio-feedback
8595 Stop forward growth if the best edge do have probability lower than this
8598 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8599 compilation for profile feedback and one for compilation without. The value
8600 for compilation with profile feedback needs to be more conservative (higher) in
8601 order to make tracer effective.
8603 @item max-cse-path-length
8605 Maximum number of basic blocks on path that cse considers. The default is 10.
8608 The maximum instructions CSE process before flushing. The default is 1000.
8610 @item ggc-min-expand
8612 GCC uses a garbage collector to manage its own memory allocation. This
8613 parameter specifies the minimum percentage by which the garbage
8614 collector's heap should be allowed to expand between collections.
8615 Tuning this may improve compilation speed; it has no effect on code
8618 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8619 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8620 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8621 GCC is not able to calculate RAM on a particular platform, the lower
8622 bound of 30% is used. Setting this parameter and
8623 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8624 every opportunity. This is extremely slow, but can be useful for
8627 @item ggc-min-heapsize
8629 Minimum size of the garbage collector's heap before it begins bothering
8630 to collect garbage. The first collection occurs after the heap expands
8631 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8632 tuning this may improve compilation speed, and has no effect on code
8635 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8636 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8637 with a lower bound of 4096 (four megabytes) and an upper bound of
8638 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8639 particular platform, the lower bound is used. Setting this parameter
8640 very large effectively disables garbage collection. Setting this
8641 parameter and @option{ggc-min-expand} to zero causes a full collection
8642 to occur at every opportunity.
8644 @item max-reload-search-insns
8645 The maximum number of instruction reload should look backward for equivalent
8646 register. Increasing values mean more aggressive optimization, making the
8647 compile time increase with probably slightly better performance. The default
8650 @item max-cselib-memory-locations
8651 The maximum number of memory locations cselib should take into account.
8652 Increasing values mean more aggressive optimization, making the compile time
8653 increase with probably slightly better performance. The default value is 500.
8655 @item reorder-blocks-duplicate
8656 @itemx reorder-blocks-duplicate-feedback
8658 Used by basic block reordering pass to decide whether to use unconditional
8659 branch or duplicate the code on its destination. Code is duplicated when its
8660 estimated size is smaller than this value multiplied by the estimated size of
8661 unconditional jump in the hot spots of the program.
8663 The @option{reorder-block-duplicate-feedback} is used only when profile
8664 feedback is available and may be set to higher values than
8665 @option{reorder-block-duplicate} since information about the hot spots is more
8668 @item max-sched-ready-insns
8669 The maximum number of instructions ready to be issued the scheduler should
8670 consider at any given time during the first scheduling pass. Increasing
8671 values mean more thorough searches, making the compilation time increase
8672 with probably little benefit. The default value is 100.
8674 @item max-sched-region-blocks
8675 The maximum number of blocks in a region to be considered for
8676 interblock scheduling. The default value is 10.
8678 @item max-pipeline-region-blocks
8679 The maximum number of blocks in a region to be considered for
8680 pipelining in the selective scheduler. The default value is 15.
8682 @item max-sched-region-insns
8683 The maximum number of insns in a region to be considered for
8684 interblock scheduling. The default value is 100.
8686 @item max-pipeline-region-insns
8687 The maximum number of insns in a region to be considered for
8688 pipelining in the selective scheduler. The default value is 200.
8691 The minimum probability (in percents) of reaching a source block
8692 for interblock speculative scheduling. The default value is 40.
8694 @item max-sched-extend-regions-iters
8695 The maximum number of iterations through CFG to extend regions.
8696 0 - disable region extension,
8697 N - do at most N iterations.
8698 The default value is 0.
8700 @item max-sched-insn-conflict-delay
8701 The maximum conflict delay for an insn to be considered for speculative motion.
8702 The default value is 3.
8704 @item sched-spec-prob-cutoff
8705 The minimal probability of speculation success (in percents), so that
8706 speculative insn will be scheduled.
8707 The default value is 40.
8709 @item sched-mem-true-dep-cost
8710 Minimal distance (in CPU cycles) between store and load targeting same
8711 memory locations. The default value is 1.
8713 @item selsched-max-lookahead
8714 The maximum size of the lookahead window of selective scheduling. It is a
8715 depth of search for available instructions.
8716 The default value is 50.
8718 @item selsched-max-sched-times
8719 The maximum number of times that an instruction will be scheduled during
8720 selective scheduling. This is the limit on the number of iterations
8721 through which the instruction may be pipelined. The default value is 2.
8723 @item selsched-max-insns-to-rename
8724 The maximum number of best instructions in the ready list that are considered
8725 for renaming in the selective scheduler. The default value is 2.
8727 @item max-last-value-rtl
8728 The maximum size measured as number of RTLs that can be recorded in an expression
8729 in combiner for a pseudo register as last known value of that register. The default
8732 @item integer-share-limit
8733 Small integer constants can use a shared data structure, reducing the
8734 compiler's memory usage and increasing its speed. This sets the maximum
8735 value of a shared integer constant. The default value is 256.
8737 @item min-virtual-mappings
8738 Specifies the minimum number of virtual mappings in the incremental
8739 SSA updater that should be registered to trigger the virtual mappings
8740 heuristic defined by virtual-mappings-ratio. The default value is
8743 @item virtual-mappings-ratio
8744 If the number of virtual mappings is virtual-mappings-ratio bigger
8745 than the number of virtual symbols to be updated, then the incremental
8746 SSA updater switches to a full update for those symbols. The default
8749 @item ssp-buffer-size
8750 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8751 protection when @option{-fstack-protection} is used.
8753 @item max-jump-thread-duplication-stmts
8754 Maximum number of statements allowed in a block that needs to be
8755 duplicated when threading jumps.
8757 @item max-fields-for-field-sensitive
8758 Maximum number of fields in a structure we will treat in
8759 a field sensitive manner during pointer analysis. The default is zero
8760 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8762 @item prefetch-latency
8763 Estimate on average number of instructions that are executed before
8764 prefetch finishes. The distance we prefetch ahead is proportional
8765 to this constant. Increasing this number may also lead to less
8766 streams being prefetched (see @option{simultaneous-prefetches}).
8768 @item simultaneous-prefetches
8769 Maximum number of prefetches that can run at the same time.
8771 @item l1-cache-line-size
8772 The size of cache line in L1 cache, in bytes.
8775 The size of L1 cache, in kilobytes.
8778 The size of L2 cache, in kilobytes.
8780 @item min-insn-to-prefetch-ratio
8781 The minimum ratio between the number of instructions and the
8782 number of prefetches to enable prefetching in a loop.
8784 @item prefetch-min-insn-to-mem-ratio
8785 The minimum ratio between the number of instructions and the
8786 number of memory references to enable prefetching in a loop.
8788 @item use-canonical-types
8789 Whether the compiler should use the ``canonical'' type system. By
8790 default, this should always be 1, which uses a more efficient internal
8791 mechanism for comparing types in C++ and Objective-C++. However, if
8792 bugs in the canonical type system are causing compilation failures,
8793 set this value to 0 to disable canonical types.
8795 @item switch-conversion-max-branch-ratio
8796 Switch initialization conversion will refuse to create arrays that are
8797 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8798 branches in the switch.
8800 @item max-partial-antic-length
8801 Maximum length of the partial antic set computed during the tree
8802 partial redundancy elimination optimization (@option{-ftree-pre}) when
8803 optimizing at @option{-O3} and above. For some sorts of source code
8804 the enhanced partial redundancy elimination optimization can run away,
8805 consuming all of the memory available on the host machine. This
8806 parameter sets a limit on the length of the sets that are computed,
8807 which prevents the runaway behavior. Setting a value of 0 for
8808 this parameter will allow an unlimited set length.
8810 @item sccvn-max-scc-size
8811 Maximum size of a strongly connected component (SCC) during SCCVN
8812 processing. If this limit is hit, SCCVN processing for the whole
8813 function will not be done and optimizations depending on it will
8814 be disabled. The default maximum SCC size is 10000.
8816 @item ira-max-loops-num
8817 IRA uses a regional register allocation by default. If a function
8818 contains loops more than number given by the parameter, only at most
8819 given number of the most frequently executed loops will form regions
8820 for the regional register allocation. The default value of the
8823 @item ira-max-conflict-table-size
8824 Although IRA uses a sophisticated algorithm of compression conflict
8825 table, the table can be still big for huge functions. If the conflict
8826 table for a function could be more than size in MB given by the
8827 parameter, the conflict table is not built and faster, simpler, and
8828 lower quality register allocation algorithm will be used. The
8829 algorithm do not use pseudo-register conflicts. The default value of
8830 the parameter is 2000.
8832 @item ira-loop-reserved-regs
8833 IRA can be used to evaluate more accurate register pressure in loops
8834 for decision to move loop invariants (see @option{-O3}). The number
8835 of available registers reserved for some other purposes is described
8836 by this parameter. The default value of the parameter is 2 which is
8837 minimal number of registers needed for execution of typical
8838 instruction. This value is the best found from numerous experiments.
8840 @item loop-invariant-max-bbs-in-loop
8841 Loop invariant motion can be very expensive, both in compile time and
8842 in amount of needed compile time memory, with very large loops. Loops
8843 with more basic blocks than this parameter won't have loop invariant
8844 motion optimization performed on them. The default value of the
8845 parameter is 1000 for -O1 and 10000 for -O2 and above.
8847 @item max-vartrack-size
8848 Sets a maximum number of hash table slots to use during variable
8849 tracking dataflow analysis of any function. If this limit is exceeded
8850 with variable tracking at assignments enabled, analysis for that
8851 function is retried without it, after removing all debug insns from
8852 the function. If the limit is exceeded even without debug insns, var
8853 tracking analysis is completely disabled for the function. Setting
8854 the parameter to zero makes it unlimited.
8856 @item min-nondebug-insn-uid
8857 Use uids starting at this parameter for nondebug insns. The range below
8858 the parameter is reserved exclusively for debug insns created by
8859 @option{-fvar-tracking-assignments}, but debug insns may get
8860 (non-overlapping) uids above it if the reserved range is exhausted.
8862 @item ipa-sra-ptr-growth-factor
8863 IPA-SRA will replace a pointer to an aggregate with one or more new
8864 parameters only when their cumulative size is less or equal to
8865 @option{ipa-sra-ptr-growth-factor} times the size of the original
8868 @item graphite-max-nb-scop-params
8869 To avoid exponential effects in the Graphite loop transforms, the
8870 number of parameters in a Static Control Part (SCoP) is bounded. The
8871 default value is 10 parameters. A variable whose value is unknown at
8872 compile time and defined outside a SCoP is a parameter of the SCoP.
8874 @item graphite-max-bbs-per-function
8875 To avoid exponential effects in the detection of SCoPs, the size of
8876 the functions analyzed by Graphite is bounded. The default value is
8879 @item loop-block-tile-size
8880 Loop blocking or strip mining transforms, enabled with
8881 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
8882 loop in the loop nest by a given number of iterations. The strip
8883 length can be changed using the @option{loop-block-tile-size}
8884 parameter. The default value is 51 iterations.
8886 @item devirt-type-list-size
8887 IPA-CP attempts to track all possible types passed to a function's
8888 parameter in order to perform devirtualization.
8889 @option{devirt-type-list-size} is the maximum number of types it
8890 stores per a single formal parameter of a function.
8892 @item lto-partitions
8893 Specify desired number of partitions produced during WHOPR compilation.
8894 The number of partitions should exceed the number of CPUs used for compilation.
8895 The default value is 32.
8897 @item lto-minpartition
8898 Size of minimal partition for WHOPR (in estimated instructions).
8899 This prevents expenses of splitting very small programs into too many
8902 @item cxx-max-namespaces-for-diagnostic-help
8903 The maximum number of namespaces to consult for suggestions when C++
8904 name lookup fails for an identifier. The default is 1000.
8909 @node Preprocessor Options
8910 @section Options Controlling the Preprocessor
8911 @cindex preprocessor options
8912 @cindex options, preprocessor
8914 These options control the C preprocessor, which is run on each C source
8915 file before actual compilation.
8917 If you use the @option{-E} option, nothing is done except preprocessing.
8918 Some of these options make sense only together with @option{-E} because
8919 they cause the preprocessor output to be unsuitable for actual
8923 @item -Wp,@var{option}
8925 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8926 and pass @var{option} directly through to the preprocessor. If
8927 @var{option} contains commas, it is split into multiple options at the
8928 commas. However, many options are modified, translated or interpreted
8929 by the compiler driver before being passed to the preprocessor, and
8930 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8931 interface is undocumented and subject to change, so whenever possible
8932 you should avoid using @option{-Wp} and let the driver handle the
8935 @item -Xpreprocessor @var{option}
8936 @opindex Xpreprocessor
8937 Pass @var{option} as an option to the preprocessor. You can use this to
8938 supply system-specific preprocessor options which GCC does not know how to
8941 If you want to pass an option that takes an argument, you must use
8942 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8945 @include cppopts.texi
8947 @node Assembler Options
8948 @section Passing Options to the Assembler
8950 @c prevent bad page break with this line
8951 You can pass options to the assembler.
8954 @item -Wa,@var{option}
8956 Pass @var{option} as an option to the assembler. If @var{option}
8957 contains commas, it is split into multiple options at the commas.
8959 @item -Xassembler @var{option}
8961 Pass @var{option} as an option to the assembler. You can use this to
8962 supply system-specific assembler options which GCC does not know how to
8965 If you want to pass an option that takes an argument, you must use
8966 @option{-Xassembler} twice, once for the option and once for the argument.
8971 @section Options for Linking
8972 @cindex link options
8973 @cindex options, linking
8975 These options come into play when the compiler links object files into
8976 an executable output file. They are meaningless if the compiler is
8977 not doing a link step.
8981 @item @var{object-file-name}
8982 A file name that does not end in a special recognized suffix is
8983 considered to name an object file or library. (Object files are
8984 distinguished from libraries by the linker according to the file
8985 contents.) If linking is done, these object files are used as input
8994 If any of these options is used, then the linker is not run, and
8995 object file names should not be used as arguments. @xref{Overall
8999 @item -l@var{library}
9000 @itemx -l @var{library}
9002 Search the library named @var{library} when linking. (The second
9003 alternative with the library as a separate argument is only for
9004 POSIX compliance and is not recommended.)
9006 It makes a difference where in the command you write this option; the
9007 linker searches and processes libraries and object files in the order they
9008 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
9009 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
9010 to functions in @samp{z}, those functions may not be loaded.
9012 The linker searches a standard list of directories for the library,
9013 which is actually a file named @file{lib@var{library}.a}. The linker
9014 then uses this file as if it had been specified precisely by name.
9016 The directories searched include several standard system directories
9017 plus any that you specify with @option{-L}.
9019 Normally the files found this way are library files---archive files
9020 whose members are object files. The linker handles an archive file by
9021 scanning through it for members which define symbols that have so far
9022 been referenced but not defined. But if the file that is found is an
9023 ordinary object file, it is linked in the usual fashion. The only
9024 difference between using an @option{-l} option and specifying a file name
9025 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
9026 and searches several directories.
9030 You need this special case of the @option{-l} option in order to
9031 link an Objective-C or Objective-C++ program.
9034 @opindex nostartfiles
9035 Do not use the standard system startup files when linking.
9036 The standard system libraries are used normally, unless @option{-nostdlib}
9037 or @option{-nodefaultlibs} is used.
9039 @item -nodefaultlibs
9040 @opindex nodefaultlibs
9041 Do not use the standard system libraries when linking.
9042 Only the libraries you specify will be passed to the linker, options
9043 specifying linkage of the system libraries, such as @code{-static-libgcc}
9044 or @code{-shared-libgcc}, will be ignored.
9045 The standard startup files are used normally, unless @option{-nostartfiles}
9046 is used. The compiler may generate calls to @code{memcmp},
9047 @code{memset}, @code{memcpy} and @code{memmove}.
9048 These entries are usually resolved by entries in
9049 libc. These entry points should be supplied through some other
9050 mechanism when this option is specified.
9054 Do not use the standard system startup files or libraries when linking.
9055 No startup files and only the libraries you specify will be passed to
9056 the linker, options specifying linkage of the system libraries, such as
9057 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
9058 The compiler may generate calls to @code{memcmp}, @code{memset},
9059 @code{memcpy} and @code{memmove}.
9060 These entries are usually resolved by entries in
9061 libc. These entry points should be supplied through some other
9062 mechanism when this option is specified.
9064 @cindex @option{-lgcc}, use with @option{-nostdlib}
9065 @cindex @option{-nostdlib} and unresolved references
9066 @cindex unresolved references and @option{-nostdlib}
9067 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
9068 @cindex @option{-nodefaultlibs} and unresolved references
9069 @cindex unresolved references and @option{-nodefaultlibs}
9070 One of the standard libraries bypassed by @option{-nostdlib} and
9071 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
9072 that GCC uses to overcome shortcomings of particular machines, or special
9073 needs for some languages.
9074 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
9075 Collection (GCC) Internals},
9076 for more discussion of @file{libgcc.a}.)
9077 In most cases, you need @file{libgcc.a} even when you want to avoid
9078 other standard libraries. In other words, when you specify @option{-nostdlib}
9079 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9080 This ensures that you have no unresolved references to internal GCC
9081 library subroutines. (For example, @samp{__main}, used to ensure C++
9082 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
9083 GNU Compiler Collection (GCC) Internals}.)
9087 Produce a position independent executable on targets which support it.
9088 For predictable results, you must also specify the same set of options
9089 that were used to generate code (@option{-fpie}, @option{-fPIE},
9090 or model suboptions) when you specify this option.
9094 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9095 that support it. This instructs the linker to add all symbols, not
9096 only used ones, to the dynamic symbol table. This option is needed
9097 for some uses of @code{dlopen} or to allow obtaining backtraces
9098 from within a program.
9102 Remove all symbol table and relocation information from the executable.
9106 On systems that support dynamic linking, this prevents linking with the shared
9107 libraries. On other systems, this option has no effect.
9111 Produce a shared object which can then be linked with other objects to
9112 form an executable. Not all systems support this option. For predictable
9113 results, you must also specify the same set of options that were used to
9114 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9115 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9116 needs to build supplementary stub code for constructors to work. On
9117 multi-libbed systems, @samp{gcc -shared} must select the correct support
9118 libraries to link against. Failing to supply the correct flags may lead
9119 to subtle defects. Supplying them in cases where they are not necessary
9122 @item -shared-libgcc
9123 @itemx -static-libgcc
9124 @opindex shared-libgcc
9125 @opindex static-libgcc
9126 On systems that provide @file{libgcc} as a shared library, these options
9127 force the use of either the shared or static version respectively.
9128 If no shared version of @file{libgcc} was built when the compiler was
9129 configured, these options have no effect.
9131 There are several situations in which an application should use the
9132 shared @file{libgcc} instead of the static version. The most common
9133 of these is when the application wishes to throw and catch exceptions
9134 across different shared libraries. In that case, each of the libraries
9135 as well as the application itself should use the shared @file{libgcc}.
9137 Therefore, the G++ and GCJ drivers automatically add
9138 @option{-shared-libgcc} whenever you build a shared library or a main
9139 executable, because C++ and Java programs typically use exceptions, so
9140 this is the right thing to do.
9142 If, instead, you use the GCC driver to create shared libraries, you may
9143 find that they will not always be linked with the shared @file{libgcc}.
9144 If GCC finds, at its configuration time, that you have a non-GNU linker
9145 or a GNU linker that does not support option @option{--eh-frame-hdr},
9146 it will link the shared version of @file{libgcc} into shared libraries
9147 by default. Otherwise, it will take advantage of the linker and optimize
9148 away the linking with the shared version of @file{libgcc}, linking with
9149 the static version of libgcc by default. This allows exceptions to
9150 propagate through such shared libraries, without incurring relocation
9151 costs at library load time.
9153 However, if a library or main executable is supposed to throw or catch
9154 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9155 for the languages used in the program, or using the option
9156 @option{-shared-libgcc}, such that it is linked with the shared
9159 @item -static-libstdc++
9160 When the @command{g++} program is used to link a C++ program, it will
9161 normally automatically link against @option{libstdc++}. If
9162 @file{libstdc++} is available as a shared library, and the
9163 @option{-static} option is not used, then this will link against the
9164 shared version of @file{libstdc++}. That is normally fine. However, it
9165 is sometimes useful to freeze the version of @file{libstdc++} used by
9166 the program without going all the way to a fully static link. The
9167 @option{-static-libstdc++} option directs the @command{g++} driver to
9168 link @file{libstdc++} statically, without necessarily linking other
9169 libraries statically.
9173 Bind references to global symbols when building a shared object. Warn
9174 about any unresolved references (unless overridden by the link editor
9175 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9178 @item -T @var{script}
9180 @cindex linker script
9181 Use @var{script} as the linker script. This option is supported by most
9182 systems using the GNU linker. On some targets, such as bare-board
9183 targets without an operating system, the @option{-T} option may be required
9184 when linking to avoid references to undefined symbols.
9186 @item -Xlinker @var{option}
9188 Pass @var{option} as an option to the linker. You can use this to
9189 supply system-specific linker options which GCC does not know how to
9192 If you want to pass an option that takes a separate argument, you must use
9193 @option{-Xlinker} twice, once for the option and once for the argument.
9194 For example, to pass @option{-assert definitions}, you must write
9195 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9196 @option{-Xlinker "-assert definitions"}, because this passes the entire
9197 string as a single argument, which is not what the linker expects.
9199 When using the GNU linker, it is usually more convenient to pass
9200 arguments to linker options using the @option{@var{option}=@var{value}}
9201 syntax than as separate arguments. For example, you can specify
9202 @samp{-Xlinker -Map=output.map} rather than
9203 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9204 this syntax for command-line options.
9206 @item -Wl,@var{option}
9208 Pass @var{option} as an option to the linker. If @var{option} contains
9209 commas, it is split into multiple options at the commas. You can use this
9210 syntax to pass an argument to the option.
9211 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9212 linker. When using the GNU linker, you can also get the same effect with
9213 @samp{-Wl,-Map=output.map}.
9215 @item -u @var{symbol}
9217 Pretend the symbol @var{symbol} is undefined, to force linking of
9218 library modules to define it. You can use @option{-u} multiple times with
9219 different symbols to force loading of additional library modules.
9222 @node Directory Options
9223 @section Options for Directory Search
9224 @cindex directory options
9225 @cindex options, directory search
9228 These options specify directories to search for header files, for
9229 libraries and for parts of the compiler:
9234 Add the directory @var{dir} to the head of the list of directories to be
9235 searched for header files. This can be used to override a system header
9236 file, substituting your own version, since these directories are
9237 searched before the system header file directories. However, you should
9238 not use this option to add directories that contain vendor-supplied
9239 system header files (use @option{-isystem} for that). If you use more than
9240 one @option{-I} option, the directories are scanned in left-to-right
9241 order; the standard system directories come after.
9243 If a standard system include directory, or a directory specified with
9244 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9245 option will be ignored. The directory will still be searched but as a
9246 system directory at its normal position in the system include chain.
9247 This is to ensure that GCC's procedure to fix buggy system headers and
9248 the ordering for the include_next directive are not inadvertently changed.
9249 If you really need to change the search order for system directories,
9250 use the @option{-nostdinc} and/or @option{-isystem} options.
9252 @item -iplugindir=@var{dir}
9253 Set the directory to search for plugins which are passed
9254 by @option{-fplugin=@var{name}} instead of
9255 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9256 to be used by the user, but only passed by the driver.
9258 @item -iquote@var{dir}
9260 Add the directory @var{dir} to the head of the list of directories to
9261 be searched for header files only for the case of @samp{#include
9262 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9263 otherwise just like @option{-I}.
9267 Add directory @var{dir} to the list of directories to be searched
9270 @item -B@var{prefix}
9272 This option specifies where to find the executables, libraries,
9273 include files, and data files of the compiler itself.
9275 The compiler driver program runs one or more of the subprograms
9276 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9277 @var{prefix} as a prefix for each program it tries to run, both with and
9278 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9280 For each subprogram to be run, the compiler driver first tries the
9281 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9282 was not specified, the driver tries two standard prefixes, which are
9283 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9284 those results in a file name that is found, the unmodified program
9285 name is searched for using the directories specified in your
9286 @env{PATH} environment variable.
9288 The compiler will check to see if the path provided by the @option{-B}
9289 refers to a directory, and if necessary it will add a directory
9290 separator character at the end of the path.
9292 @option{-B} prefixes that effectively specify directory names also apply
9293 to libraries in the linker, because the compiler translates these
9294 options into @option{-L} options for the linker. They also apply to
9295 includes files in the preprocessor, because the compiler translates these
9296 options into @option{-isystem} options for the preprocessor. In this case,
9297 the compiler appends @samp{include} to the prefix.
9299 The run-time support file @file{libgcc.a} can also be searched for using
9300 the @option{-B} prefix, if needed. If it is not found there, the two
9301 standard prefixes above are tried, and that is all. The file is left
9302 out of the link if it is not found by those means.
9304 Another way to specify a prefix much like the @option{-B} prefix is to use
9305 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9308 As a special kludge, if the path provided by @option{-B} is
9309 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9310 9, then it will be replaced by @file{[dir/]include}. This is to help
9311 with boot-strapping the compiler.
9313 @item -specs=@var{file}
9315 Process @var{file} after the compiler reads in the standard @file{specs}
9316 file, in order to override the defaults that the @file{gcc} driver
9317 program uses when determining what switches to pass to @file{cc1},
9318 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9319 @option{-specs=@var{file}} can be specified on the command line, and they
9320 are processed in order, from left to right.
9322 @item --sysroot=@var{dir}
9324 Use @var{dir} as the logical root directory for headers and libraries.
9325 For example, if the compiler would normally search for headers in
9326 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9327 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9329 If you use both this option and the @option{-isysroot} option, then
9330 the @option{--sysroot} option will apply to libraries, but the
9331 @option{-isysroot} option will apply to header files.
9333 The GNU linker (beginning with version 2.16) has the necessary support
9334 for this option. If your linker does not support this option, the
9335 header file aspect of @option{--sysroot} will still work, but the
9336 library aspect will not.
9340 This option has been deprecated. Please use @option{-iquote} instead for
9341 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9342 Any directories you specify with @option{-I} options before the @option{-I-}
9343 option are searched only for the case of @samp{#include "@var{file}"};
9344 they are not searched for @samp{#include <@var{file}>}.
9346 If additional directories are specified with @option{-I} options after
9347 the @option{-I-}, these directories are searched for all @samp{#include}
9348 directives. (Ordinarily @emph{all} @option{-I} directories are used
9351 In addition, the @option{-I-} option inhibits the use of the current
9352 directory (where the current input file came from) as the first search
9353 directory for @samp{#include "@var{file}"}. There is no way to
9354 override this effect of @option{-I-}. With @option{-I.} you can specify
9355 searching the directory which was current when the compiler was
9356 invoked. That is not exactly the same as what the preprocessor does
9357 by default, but it is often satisfactory.
9359 @option{-I-} does not inhibit the use of the standard system directories
9360 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9367 @section Specifying subprocesses and the switches to pass to them
9370 @command{gcc} is a driver program. It performs its job by invoking a
9371 sequence of other programs to do the work of compiling, assembling and
9372 linking. GCC interprets its command-line parameters and uses these to
9373 deduce which programs it should invoke, and which command-line options
9374 it ought to place on their command lines. This behavior is controlled
9375 by @dfn{spec strings}. In most cases there is one spec string for each
9376 program that GCC can invoke, but a few programs have multiple spec
9377 strings to control their behavior. The spec strings built into GCC can
9378 be overridden by using the @option{-specs=} command-line switch to specify
9381 @dfn{Spec files} are plaintext files that are used to construct spec
9382 strings. They consist of a sequence of directives separated by blank
9383 lines. The type of directive is determined by the first non-whitespace
9384 character on the line and it can be one of the following:
9387 @item %@var{command}
9388 Issues a @var{command} to the spec file processor. The commands that can
9392 @item %include <@var{file}>
9393 @cindex @code{%include}
9394 Search for @var{file} and insert its text at the current point in the
9397 @item %include_noerr <@var{file}>
9398 @cindex @code{%include_noerr}
9399 Just like @samp{%include}, but do not generate an error message if the include
9400 file cannot be found.
9402 @item %rename @var{old_name} @var{new_name}
9403 @cindex @code{%rename}
9404 Rename the spec string @var{old_name} to @var{new_name}.
9408 @item *[@var{spec_name}]:
9409 This tells the compiler to create, override or delete the named spec
9410 string. All lines after this directive up to the next directive or
9411 blank line are considered to be the text for the spec string. If this
9412 results in an empty string then the spec will be deleted. (Or, if the
9413 spec did not exist, then nothing will happened.) Otherwise, if the spec
9414 does not currently exist a new spec will be created. If the spec does
9415 exist then its contents will be overridden by the text of this
9416 directive, unless the first character of that text is the @samp{+}
9417 character, in which case the text will be appended to the spec.
9419 @item [@var{suffix}]:
9420 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9421 and up to the next directive or blank line are considered to make up the
9422 spec string for the indicated suffix. When the compiler encounters an
9423 input file with the named suffix, it will processes the spec string in
9424 order to work out how to compile that file. For example:
9431 This says that any input file whose name ends in @samp{.ZZ} should be
9432 passed to the program @samp{z-compile}, which should be invoked with the
9433 command-line switch @option{-input} and with the result of performing the
9434 @samp{%i} substitution. (See below.)
9436 As an alternative to providing a spec string, the text that follows a
9437 suffix directive can be one of the following:
9440 @item @@@var{language}
9441 This says that the suffix is an alias for a known @var{language}. This is
9442 similar to using the @option{-x} command-line switch to GCC to specify a
9443 language explicitly. For example:
9450 Says that .ZZ files are, in fact, C++ source files.
9453 This causes an error messages saying:
9456 @var{name} compiler not installed on this system.
9460 GCC already has an extensive list of suffixes built into it.
9461 This directive will add an entry to the end of the list of suffixes, but
9462 since the list is searched from the end backwards, it is effectively
9463 possible to override earlier entries using this technique.
9467 GCC has the following spec strings built into it. Spec files can
9468 override these strings or create their own. Note that individual
9469 targets can also add their own spec strings to this list.
9472 asm Options to pass to the assembler
9473 asm_final Options to pass to the assembler post-processor
9474 cpp Options to pass to the C preprocessor
9475 cc1 Options to pass to the C compiler
9476 cc1plus Options to pass to the C++ compiler
9477 endfile Object files to include at the end of the link
9478 link Options to pass to the linker
9479 lib Libraries to include on the command line to the linker
9480 libgcc Decides which GCC support library to pass to the linker
9481 linker Sets the name of the linker
9482 predefines Defines to be passed to the C preprocessor
9483 signed_char Defines to pass to CPP to say whether @code{char} is signed
9485 startfile Object files to include at the start of the link
9488 Here is a small example of a spec file:
9494 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9497 This example renames the spec called @samp{lib} to @samp{old_lib} and
9498 then overrides the previous definition of @samp{lib} with a new one.
9499 The new definition adds in some extra command-line options before
9500 including the text of the old definition.
9502 @dfn{Spec strings} are a list of command-line options to be passed to their
9503 corresponding program. In addition, the spec strings can contain
9504 @samp{%}-prefixed sequences to substitute variable text or to
9505 conditionally insert text into the command line. Using these constructs
9506 it is possible to generate quite complex command lines.
9508 Here is a table of all defined @samp{%}-sequences for spec
9509 strings. Note that spaces are not generated automatically around the
9510 results of expanding these sequences. Therefore you can concatenate them
9511 together or combine them with constant text in a single argument.
9515 Substitute one @samp{%} into the program name or argument.
9518 Substitute the name of the input file being processed.
9521 Substitute the basename of the input file being processed.
9522 This is the substring up to (and not including) the last period
9523 and not including the directory.
9526 This is the same as @samp{%b}, but include the file suffix (text after
9530 Marks the argument containing or following the @samp{%d} as a
9531 temporary file name, so that that file will be deleted if GCC exits
9532 successfully. Unlike @samp{%g}, this contributes no text to the
9535 @item %g@var{suffix}
9536 Substitute a file name that has suffix @var{suffix} and is chosen
9537 once per compilation, and mark the argument in the same way as
9538 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9539 name is now chosen in a way that is hard to predict even when previously
9540 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9541 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9542 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9543 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9544 was simply substituted with a file name chosen once per compilation,
9545 without regard to any appended suffix (which was therefore treated
9546 just like ordinary text), making such attacks more likely to succeed.
9548 @item %u@var{suffix}
9549 Like @samp{%g}, but generates a new temporary file name even if
9550 @samp{%u@var{suffix}} was already seen.
9552 @item %U@var{suffix}
9553 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9554 new one if there is no such last file name. In the absence of any
9555 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9556 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9557 would involve the generation of two distinct file names, one
9558 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9559 simply substituted with a file name chosen for the previous @samp{%u},
9560 without regard to any appended suffix.
9562 @item %j@var{suffix}
9563 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9564 writable, and if save-temps is off; otherwise, substitute the name
9565 of a temporary file, just like @samp{%u}. This temporary file is not
9566 meant for communication between processes, but rather as a junk
9569 @item %|@var{suffix}
9570 @itemx %m@var{suffix}
9571 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9572 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9573 all. These are the two most common ways to instruct a program that it
9574 should read from standard input or write to standard output. If you
9575 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9576 construct: see for example @file{f/lang-specs.h}.
9578 @item %.@var{SUFFIX}
9579 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9580 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9581 terminated by the next space or %.
9584 Marks the argument containing or following the @samp{%w} as the
9585 designated output file of this compilation. This puts the argument
9586 into the sequence of arguments that @samp{%o} will substitute later.
9589 Substitutes the names of all the output files, with spaces
9590 automatically placed around them. You should write spaces
9591 around the @samp{%o} as well or the results are undefined.
9592 @samp{%o} is for use in the specs for running the linker.
9593 Input files whose names have no recognized suffix are not compiled
9594 at all, but they are included among the output files, so they will
9598 Substitutes the suffix for object files. Note that this is
9599 handled specially when it immediately follows @samp{%g, %u, or %U},
9600 because of the need for those to form complete file names. The
9601 handling is such that @samp{%O} is treated exactly as if it had already
9602 been substituted, except that @samp{%g, %u, and %U} do not currently
9603 support additional @var{suffix} characters following @samp{%O} as they would
9604 following, for example, @samp{.o}.
9607 Substitutes the standard macro predefinitions for the
9608 current target machine. Use this when running @code{cpp}.
9611 Like @samp{%p}, but puts @samp{__} before and after the name of each
9612 predefined macro, except for macros that start with @samp{__} or with
9613 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9617 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9618 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9619 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9620 and @option{-imultilib} as necessary.
9623 Current argument is the name of a library or startup file of some sort.
9624 Search for that file in a standard list of directories and substitute
9625 the full name found. The current working directory is included in the
9626 list of directories scanned.
9629 Current argument is the name of a linker script. Search for that file
9630 in the current list of directories to scan for libraries. If the file
9631 is located insert a @option{--script} option into the command line
9632 followed by the full path name found. If the file is not found then
9633 generate an error message. Note: the current working directory is not
9637 Print @var{str} as an error message. @var{str} is terminated by a newline.
9638 Use this when inconsistent options are detected.
9641 Substitute the contents of spec string @var{name} at this point.
9644 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9646 @item %x@{@var{option}@}
9647 Accumulate an option for @samp{%X}.
9650 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9654 Output the accumulated assembler options specified by @option{-Wa}.
9657 Output the accumulated preprocessor options specified by @option{-Wp}.
9660 Process the @code{asm} spec. This is used to compute the
9661 switches to be passed to the assembler.
9664 Process the @code{asm_final} spec. This is a spec string for
9665 passing switches to an assembler post-processor, if such a program is
9669 Process the @code{link} spec. This is the spec for computing the
9670 command line passed to the linker. Typically it will make use of the
9671 @samp{%L %G %S %D and %E} sequences.
9674 Dump out a @option{-L} option for each directory that GCC believes might
9675 contain startup files. If the target supports multilibs then the
9676 current multilib directory will be prepended to each of these paths.
9679 Process the @code{lib} spec. This is a spec string for deciding which
9680 libraries should be included on the command line to the linker.
9683 Process the @code{libgcc} spec. This is a spec string for deciding
9684 which GCC support library should be included on the command line to the linker.
9687 Process the @code{startfile} spec. This is a spec for deciding which
9688 object files should be the first ones passed to the linker. Typically
9689 this might be a file named @file{crt0.o}.
9692 Process the @code{endfile} spec. This is a spec string that specifies
9693 the last object files that will be passed to the linker.
9696 Process the @code{cpp} spec. This is used to construct the arguments
9697 to be passed to the C preprocessor.
9700 Process the @code{cc1} spec. This is used to construct the options to be
9701 passed to the actual C compiler (@samp{cc1}).
9704 Process the @code{cc1plus} spec. This is used to construct the options to be
9705 passed to the actual C++ compiler (@samp{cc1plus}).
9708 Substitute the variable part of a matched option. See below.
9709 Note that each comma in the substituted string is replaced by
9713 Remove all occurrences of @code{-S} from the command line. Note---this
9714 command is position dependent. @samp{%} commands in the spec string
9715 before this one will see @code{-S}, @samp{%} commands in the spec string
9716 after this one will not.
9718 @item %:@var{function}(@var{args})
9719 Call the named function @var{function}, passing it @var{args}.
9720 @var{args} is first processed as a nested spec string, then split
9721 into an argument vector in the usual fashion. The function returns
9722 a string which is processed as if it had appeared literally as part
9723 of the current spec.
9725 The following built-in spec functions are provided:
9729 The @code{getenv} spec function takes two arguments: an environment
9730 variable name and a string. If the environment variable is not
9731 defined, a fatal error is issued. Otherwise, the return value is the
9732 value of the environment variable concatenated with the string. For
9733 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9736 %:getenv(TOPDIR /include)
9739 expands to @file{/path/to/top/include}.
9741 @item @code{if-exists}
9742 The @code{if-exists} spec function takes one argument, an absolute
9743 pathname to a file. If the file exists, @code{if-exists} returns the
9744 pathname. Here is a small example of its usage:
9748 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9751 @item @code{if-exists-else}
9752 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9753 spec function, except that it takes two arguments. The first argument is
9754 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9755 returns the pathname. If it does not exist, it returns the second argument.
9756 This way, @code{if-exists-else} can be used to select one file or another,
9757 based on the existence of the first. Here is a small example of its usage:
9761 crt0%O%s %:if-exists(crti%O%s) \
9762 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9765 @item @code{replace-outfile}
9766 The @code{replace-outfile} spec function takes two arguments. It looks for the
9767 first argument in the outfiles array and replaces it with the second argument. Here
9768 is a small example of its usage:
9771 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9774 @item @code{remove-outfile}
9775 The @code{remove-outfile} spec function takes one argument. It looks for the
9776 first argument in the outfiles array and removes it. Here is a small example
9780 %:remove-outfile(-lm)
9783 @item @code{pass-through-libs}
9784 The @code{pass-through-libs} spec function takes any number of arguments. It
9785 finds any @option{-l} options and any non-options ending in ".a" (which it
9786 assumes are the names of linker input library archive files) and returns a
9787 result containing all the found arguments each prepended by
9788 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
9789 intended to be passed to the LTO linker plugin.
9792 %:pass-through-libs(%G %L %G)
9795 @item @code{print-asm-header}
9796 The @code{print-asm-header} function takes no arguments and simply
9797 prints a banner like:
9803 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9806 It is used to separate compiler options from assembler options
9807 in the @option{--target-help} output.
9811 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9812 If that switch was not specified, this substitutes nothing. Note that
9813 the leading dash is omitted when specifying this option, and it is
9814 automatically inserted if the substitution is performed. Thus the spec
9815 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9816 and would output the command line option @option{-foo}.
9818 @item %W@{@code{S}@}
9819 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9822 @item %@{@code{S}*@}
9823 Substitutes all the switches specified to GCC whose names start
9824 with @code{-S}, but which also take an argument. This is used for
9825 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9826 GCC considers @option{-o foo} as being
9827 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9828 text, including the space. Thus two arguments would be generated.
9830 @item %@{@code{S}*&@code{T}*@}
9831 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9832 (the order of @code{S} and @code{T} in the spec is not significant).
9833 There can be any number of ampersand-separated variables; for each the
9834 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9836 @item %@{@code{S}:@code{X}@}
9837 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9839 @item %@{!@code{S}:@code{X}@}
9840 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9842 @item %@{@code{S}*:@code{X}@}
9843 Substitutes @code{X} if one or more switches whose names start with
9844 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9845 once, no matter how many such switches appeared. However, if @code{%*}
9846 appears somewhere in @code{X}, then @code{X} will be substituted once
9847 for each matching switch, with the @code{%*} replaced by the part of
9848 that switch that matched the @code{*}.
9850 @item %@{.@code{S}:@code{X}@}
9851 Substitutes @code{X}, if processing a file with suffix @code{S}.
9853 @item %@{!.@code{S}:@code{X}@}
9854 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9856 @item %@{,@code{S}:@code{X}@}
9857 Substitutes @code{X}, if processing a file for language @code{S}.
9859 @item %@{!,@code{S}:@code{X}@}
9860 Substitutes @code{X}, if not processing a file for language @code{S}.
9862 @item %@{@code{S}|@code{P}:@code{X}@}
9863 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9864 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9865 @code{*} sequences as well, although they have a stronger binding than
9866 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9867 alternatives must be starred, and only the first matching alternative
9870 For example, a spec string like this:
9873 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9876 will output the following command-line options from the following input
9877 command-line options:
9882 -d fred.c -foo -baz -boggle
9883 -d jim.d -bar -baz -boggle
9886 @item %@{S:X; T:Y; :D@}
9888 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9889 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9890 be as many clauses as you need. This may be combined with @code{.},
9891 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9896 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9897 construct may contain other nested @samp{%} constructs or spaces, or
9898 even newlines. They are processed as usual, as described above.
9899 Trailing white space in @code{X} is ignored. White space may also
9900 appear anywhere on the left side of the colon in these constructs,
9901 except between @code{.} or @code{*} and the corresponding word.
9903 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9904 handled specifically in these constructs. If another value of
9905 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9906 @option{-W} switch is found later in the command line, the earlier
9907 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9908 just one letter, which passes all matching options.
9910 The character @samp{|} at the beginning of the predicate text is used to
9911 indicate that a command should be piped to the following command, but
9912 only if @option{-pipe} is specified.
9914 It is built into GCC which switches take arguments and which do not.
9915 (You might think it would be useful to generalize this to allow each
9916 compiler's spec to say which switches take arguments. But this cannot
9917 be done in a consistent fashion. GCC cannot even decide which input
9918 files have been specified without knowing which switches take arguments,
9919 and it must know which input files to compile in order to tell which
9922 GCC also knows implicitly that arguments starting in @option{-l} are to be
9923 treated as compiler output files, and passed to the linker in their
9924 proper position among the other output files.
9926 @c man begin OPTIONS
9928 @node Target Options
9929 @section Specifying Target Machine and Compiler Version
9930 @cindex target options
9931 @cindex cross compiling
9932 @cindex specifying machine version
9933 @cindex specifying compiler version and target machine
9934 @cindex compiler version, specifying
9935 @cindex target machine, specifying
9937 The usual way to run GCC is to run the executable called @command{gcc}, or
9938 @command{@var{machine}-gcc} when cross-compiling, or
9939 @command{@var{machine}-gcc-@var{version}} to run a version other than the
9940 one that was installed last.
9942 @node Submodel Options
9943 @section Hardware Models and Configurations
9944 @cindex submodel options
9945 @cindex specifying hardware config
9946 @cindex hardware models and configurations, specifying
9947 @cindex machine dependent options
9949 Each target machine types can have its own
9950 special options, starting with @samp{-m}, to choose among various
9951 hardware models or configurations---for example, 68010 vs 68020,
9952 floating coprocessor or none. A single installed version of the
9953 compiler can compile for any model or configuration, according to the
9956 Some configurations of the compiler also support additional special
9957 options, usually for compatibility with other compilers on the same
9960 @c This list is ordered alphanumerically by subsection name.
9961 @c It should be the same order and spelling as these options are listed
9962 @c in Machine Dependent Options
9968 * Blackfin Options::
9972 * DEC Alpha Options::
9973 * DEC Alpha/VMS Options::
9976 * GNU/Linux Options::
9979 * i386 and x86-64 Options::
9980 * i386 and x86-64 Windows Options::
9982 * IA-64/VMS Options::
9990 * MicroBlaze Options::
9995 * picoChip Options::
9997 * RS/6000 and PowerPC Options::
9999 * S/390 and zSeries Options::
10002 * Solaris 2 Options::
10005 * System V Options::
10008 * VxWorks Options::
10010 * Xstormy16 Options::
10012 * zSeries Options::
10016 @subsection ARC Options
10017 @cindex ARC Options
10019 These options are defined for ARC implementations:
10024 Compile code for little endian mode. This is the default.
10028 Compile code for big endian mode.
10031 @opindex mmangle-cpu
10032 Prepend the name of the CPU to all public symbol names.
10033 In multiple-processor systems, there are many ARC variants with different
10034 instruction and register set characteristics. This flag prevents code
10035 compiled for one CPU to be linked with code compiled for another.
10036 No facility exists for handling variants that are ``almost identical''.
10037 This is an all or nothing option.
10039 @item -mcpu=@var{cpu}
10041 Compile code for ARC variant @var{cpu}.
10042 Which variants are supported depend on the configuration.
10043 All variants support @option{-mcpu=base}, this is the default.
10045 @item -mtext=@var{text-section}
10046 @itemx -mdata=@var{data-section}
10047 @itemx -mrodata=@var{readonly-data-section}
10051 Put functions, data, and readonly data in @var{text-section},
10052 @var{data-section}, and @var{readonly-data-section} respectively
10053 by default. This can be overridden with the @code{section} attribute.
10054 @xref{Variable Attributes}.
10059 @subsection ARM Options
10060 @cindex ARM options
10062 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
10066 @item -mabi=@var{name}
10068 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
10069 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
10072 @opindex mapcs-frame
10073 Generate a stack frame that is compliant with the ARM Procedure Call
10074 Standard for all functions, even if this is not strictly necessary for
10075 correct execution of the code. Specifying @option{-fomit-frame-pointer}
10076 with this option will cause the stack frames not to be generated for
10077 leaf functions. The default is @option{-mno-apcs-frame}.
10081 This is a synonym for @option{-mapcs-frame}.
10084 @c not currently implemented
10085 @item -mapcs-stack-check
10086 @opindex mapcs-stack-check
10087 Generate code to check the amount of stack space available upon entry to
10088 every function (that actually uses some stack space). If there is
10089 insufficient space available then either the function
10090 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10091 called, depending upon the amount of stack space required. The run time
10092 system is required to provide these functions. The default is
10093 @option{-mno-apcs-stack-check}, since this produces smaller code.
10095 @c not currently implemented
10097 @opindex mapcs-float
10098 Pass floating point arguments using the float point registers. This is
10099 one of the variants of the APCS@. This option is recommended if the
10100 target hardware has a floating point unit or if a lot of floating point
10101 arithmetic is going to be performed by the code. The default is
10102 @option{-mno-apcs-float}, since integer only code is slightly increased in
10103 size if @option{-mapcs-float} is used.
10105 @c not currently implemented
10106 @item -mapcs-reentrant
10107 @opindex mapcs-reentrant
10108 Generate reentrant, position independent code. The default is
10109 @option{-mno-apcs-reentrant}.
10112 @item -mthumb-interwork
10113 @opindex mthumb-interwork
10114 Generate code which supports calling between the ARM and Thumb
10115 instruction sets. Without this option the two instruction sets cannot
10116 be reliably used inside one program. The default is
10117 @option{-mno-thumb-interwork}, since slightly larger code is generated
10118 when @option{-mthumb-interwork} is specified.
10120 @item -mno-sched-prolog
10121 @opindex mno-sched-prolog
10122 Prevent the reordering of instructions in the function prolog, or the
10123 merging of those instruction with the instructions in the function's
10124 body. This means that all functions will start with a recognizable set
10125 of instructions (or in fact one of a choice from a small set of
10126 different function prologues), and this information can be used to
10127 locate the start if functions inside an executable piece of code. The
10128 default is @option{-msched-prolog}.
10130 @item -mfloat-abi=@var{name}
10131 @opindex mfloat-abi
10132 Specifies which floating-point ABI to use. Permissible values
10133 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10135 Specifying @samp{soft} causes GCC to generate output containing
10136 library calls for floating-point operations.
10137 @samp{softfp} allows the generation of code using hardware floating-point
10138 instructions, but still uses the soft-float calling conventions.
10139 @samp{hard} allows generation of floating-point instructions
10140 and uses FPU-specific calling conventions.
10142 The default depends on the specific target configuration. Note that
10143 the hard-float and soft-float ABIs are not link-compatible; you must
10144 compile your entire program with the same ABI, and link with a
10145 compatible set of libraries.
10148 @opindex mhard-float
10149 Equivalent to @option{-mfloat-abi=hard}.
10152 @opindex msoft-float
10153 Equivalent to @option{-mfloat-abi=soft}.
10155 @item -mlittle-endian
10156 @opindex mlittle-endian
10157 Generate code for a processor running in little-endian mode. This is
10158 the default for all standard configurations.
10161 @opindex mbig-endian
10162 Generate code for a processor running in big-endian mode; the default is
10163 to compile code for a little-endian processor.
10165 @item -mwords-little-endian
10166 @opindex mwords-little-endian
10167 This option only applies when generating code for big-endian processors.
10168 Generate code for a little-endian word order but a big-endian byte
10169 order. That is, a byte order of the form @samp{32107654}. Note: this
10170 option should only be used if you require compatibility with code for
10171 big-endian ARM processors generated by versions of the compiler prior to
10174 @item -mcpu=@var{name}
10176 This specifies the name of the target ARM processor. GCC uses this name
10177 to determine what kind of instructions it can emit when generating
10178 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10179 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10180 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10181 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10182 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10184 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10185 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10186 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10187 @samp{strongarm1110},
10188 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10189 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10190 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10191 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10192 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10193 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10194 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10195 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a15},
10196 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m4}, @samp{cortex-m3},
10199 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10201 @item -mtune=@var{name}
10203 This option is very similar to the @option{-mcpu=} option, except that
10204 instead of specifying the actual target processor type, and hence
10205 restricting which instructions can be used, it specifies that GCC should
10206 tune the performance of the code as if the target were of the type
10207 specified in this option, but still choosing the instructions that it
10208 will generate based on the CPU specified by a @option{-mcpu=} option.
10209 For some ARM implementations better performance can be obtained by using
10212 @item -march=@var{name}
10214 This specifies the name of the target ARM architecture. GCC uses this
10215 name to determine what kind of instructions it can emit when generating
10216 assembly code. This option can be used in conjunction with or instead
10217 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10218 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10219 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10220 @samp{armv6}, @samp{armv6j},
10221 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10222 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10223 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10225 @item -mfpu=@var{name}
10226 @itemx -mfpe=@var{number}
10227 @itemx -mfp=@var{number}
10231 This specifies what floating point hardware (or hardware emulation) is
10232 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10233 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10234 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10235 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10236 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10237 @option{-mfp} and @option{-mfpe} are synonyms for
10238 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10241 If @option{-msoft-float} is specified this specifies the format of
10242 floating point values.
10244 If the selected floating-point hardware includes the NEON extension
10245 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10246 operations will not be used by GCC's auto-vectorization pass unless
10247 @option{-funsafe-math-optimizations} is also specified. This is
10248 because NEON hardware does not fully implement the IEEE 754 standard for
10249 floating-point arithmetic (in particular denormal values are treated as
10250 zero), so the use of NEON instructions may lead to a loss of precision.
10252 @item -mfp16-format=@var{name}
10253 @opindex mfp16-format
10254 Specify the format of the @code{__fp16} half-precision floating-point type.
10255 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10256 the default is @samp{none}, in which case the @code{__fp16} type is not
10257 defined. @xref{Half-Precision}, for more information.
10259 @item -mstructure-size-boundary=@var{n}
10260 @opindex mstructure-size-boundary
10261 The size of all structures and unions will be rounded up to a multiple
10262 of the number of bits set by this option. Permissible values are 8, 32
10263 and 64. The default value varies for different toolchains. For the COFF
10264 targeted toolchain the default value is 8. A value of 64 is only allowed
10265 if the underlying ABI supports it.
10267 Specifying the larger number can produce faster, more efficient code, but
10268 can also increase the size of the program. Different values are potentially
10269 incompatible. Code compiled with one value cannot necessarily expect to
10270 work with code or libraries compiled with another value, if they exchange
10271 information using structures or unions.
10273 @item -mabort-on-noreturn
10274 @opindex mabort-on-noreturn
10275 Generate a call to the function @code{abort} at the end of a
10276 @code{noreturn} function. It will be executed if the function tries to
10280 @itemx -mno-long-calls
10281 @opindex mlong-calls
10282 @opindex mno-long-calls
10283 Tells the compiler to perform function calls by first loading the
10284 address of the function into a register and then performing a subroutine
10285 call on this register. This switch is needed if the target function
10286 will lie outside of the 64 megabyte addressing range of the offset based
10287 version of subroutine call instruction.
10289 Even if this switch is enabled, not all function calls will be turned
10290 into long calls. The heuristic is that static functions, functions
10291 which have the @samp{short-call} attribute, functions that are inside
10292 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10293 definitions have already been compiled within the current compilation
10294 unit, will not be turned into long calls. The exception to this rule is
10295 that weak function definitions, functions with the @samp{long-call}
10296 attribute or the @samp{section} attribute, and functions that are within
10297 the scope of a @samp{#pragma long_calls} directive, will always be
10298 turned into long calls.
10300 This feature is not enabled by default. Specifying
10301 @option{-mno-long-calls} will restore the default behavior, as will
10302 placing the function calls within the scope of a @samp{#pragma
10303 long_calls_off} directive. Note these switches have no effect on how
10304 the compiler generates code to handle function calls via function
10307 @item -msingle-pic-base
10308 @opindex msingle-pic-base
10309 Treat the register used for PIC addressing as read-only, rather than
10310 loading it in the prologue for each function. The run-time system is
10311 responsible for initializing this register with an appropriate value
10312 before execution begins.
10314 @item -mpic-register=@var{reg}
10315 @opindex mpic-register
10316 Specify the register to be used for PIC addressing. The default is R10
10317 unless stack-checking is enabled, when R9 is used.
10319 @item -mcirrus-fix-invalid-insns
10320 @opindex mcirrus-fix-invalid-insns
10321 @opindex mno-cirrus-fix-invalid-insns
10322 Insert NOPs into the instruction stream to in order to work around
10323 problems with invalid Maverick instruction combinations. This option
10324 is only valid if the @option{-mcpu=ep9312} option has been used to
10325 enable generation of instructions for the Cirrus Maverick floating
10326 point co-processor. This option is not enabled by default, since the
10327 problem is only present in older Maverick implementations. The default
10328 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10331 @item -mpoke-function-name
10332 @opindex mpoke-function-name
10333 Write the name of each function into the text section, directly
10334 preceding the function prologue. The generated code is similar to this:
10338 .ascii "arm_poke_function_name", 0
10341 .word 0xff000000 + (t1 - t0)
10342 arm_poke_function_name
10344 stmfd sp!, @{fp, ip, lr, pc@}
10348 When performing a stack backtrace, code can inspect the value of
10349 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10350 location @code{pc - 12} and the top 8 bits are set, then we know that
10351 there is a function name embedded immediately preceding this location
10352 and has length @code{((pc[-3]) & 0xff000000)}.
10356 Generate code for the Thumb instruction set. The default is to
10357 use the 32-bit ARM instruction set.
10358 This option automatically enables either 16-bit Thumb-1 or
10359 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
10360 and @option{-march=@var{name}} options. This option is not passed to the
10361 assembler. If you want to force assembler files to be interpreted as Thumb code,
10362 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
10363 option directly to the assembler by prefixing it with @option{-Wa}.
10366 @opindex mtpcs-frame
10367 Generate a stack frame that is compliant with the Thumb Procedure Call
10368 Standard for all non-leaf functions. (A leaf function is one that does
10369 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10371 @item -mtpcs-leaf-frame
10372 @opindex mtpcs-leaf-frame
10373 Generate a stack frame that is compliant with the Thumb Procedure Call
10374 Standard for all leaf functions. (A leaf function is one that does
10375 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10377 @item -mcallee-super-interworking
10378 @opindex mcallee-super-interworking
10379 Gives all externally visible functions in the file being compiled an ARM
10380 instruction set header which switches to Thumb mode before executing the
10381 rest of the function. This allows these functions to be called from
10382 non-interworking code. This option is not valid in AAPCS configurations
10383 because interworking is enabled by default.
10385 @item -mcaller-super-interworking
10386 @opindex mcaller-super-interworking
10387 Allows calls via function pointers (including virtual functions) to
10388 execute correctly regardless of whether the target code has been
10389 compiled for interworking or not. There is a small overhead in the cost
10390 of executing a function pointer if this option is enabled. This option
10391 is not valid in AAPCS configurations because interworking is enabled
10394 @item -mtp=@var{name}
10396 Specify the access model for the thread local storage pointer. The valid
10397 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10398 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10399 (supported in the arm6k architecture), and @option{auto}, which uses the
10400 best available method for the selected processor. The default setting is
10403 @item -mword-relocations
10404 @opindex mword-relocations
10405 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10406 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10407 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10410 @item -mfix-cortex-m3-ldrd
10411 @opindex mfix-cortex-m3-ldrd
10412 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10413 with overlapping destination and base registers are used. This option avoids
10414 generating these instructions. This option is enabled by default when
10415 @option{-mcpu=cortex-m3} is specified.
10420 @subsection AVR Options
10421 @cindex AVR Options
10423 These options are defined for AVR implementations:
10426 @item -mmcu=@var{mcu}
10428 Specify ATMEL AVR instruction set or MCU type.
10430 Instruction set avr1 is for the minimal AVR core, not supported by the C
10431 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10432 attiny11, attiny12, attiny15, attiny28).
10434 Instruction set avr2 (default) is for the classic AVR core with up to
10435 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10436 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10437 at90c8534, at90s8535).
10439 Instruction set avr3 is for the classic AVR core with up to 128K program
10440 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10442 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10443 memory space (MCU types: atmega8, atmega83, atmega85).
10445 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10446 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10447 atmega64, atmega128, at43usb355, at94k).
10449 @item -mno-interrupts
10450 @opindex mno-interrupts
10451 Generated code is not compatible with hardware interrupts.
10452 Code size will be smaller.
10454 @item -mcall-prologues
10455 @opindex mcall-prologues
10456 Functions prologues/epilogues expanded as call to appropriate
10457 subroutines. Code size will be smaller.
10460 @opindex mtiny-stack
10461 Change only the low 8 bits of the stack pointer.
10465 Assume int to be 8 bit integer. This affects the sizes of all types: A
10466 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10467 and long long will be 4 bytes. Please note that this option does not
10468 comply to the C standards, but it will provide you with smaller code
10472 @node Blackfin Options
10473 @subsection Blackfin Options
10474 @cindex Blackfin Options
10477 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10479 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10480 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10481 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10482 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10483 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10484 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10485 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10487 The optional @var{sirevision} specifies the silicon revision of the target
10488 Blackfin processor. Any workarounds available for the targeted silicon revision
10489 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10490 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10491 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10492 hexadecimal digits representing the major and minor numbers in the silicon
10493 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10494 is not defined. If @var{sirevision} is @samp{any}, the
10495 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10496 If this optional @var{sirevision} is not used, GCC assumes the latest known
10497 silicon revision of the targeted Blackfin processor.
10499 Support for @samp{bf561} is incomplete. For @samp{bf561},
10500 Only the processor macro is defined.
10501 Without this option, @samp{bf532} is used as the processor by default.
10502 The corresponding predefined processor macros for @var{cpu} is to
10503 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10504 provided by libgloss to be linked in if @option{-msim} is not given.
10508 Specifies that the program will be run on the simulator. This causes
10509 the simulator BSP provided by libgloss to be linked in. This option
10510 has effect only for @samp{bfin-elf} toolchain.
10511 Certain other options, such as @option{-mid-shared-library} and
10512 @option{-mfdpic}, imply @option{-msim}.
10514 @item -momit-leaf-frame-pointer
10515 @opindex momit-leaf-frame-pointer
10516 Don't keep the frame pointer in a register for leaf functions. This
10517 avoids the instructions to save, set up and restore frame pointers and
10518 makes an extra register available in leaf functions. The option
10519 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10520 which might make debugging harder.
10522 @item -mspecld-anomaly
10523 @opindex mspecld-anomaly
10524 When enabled, the compiler will ensure that the generated code does not
10525 contain speculative loads after jump instructions. If this option is used,
10526 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10528 @item -mno-specld-anomaly
10529 @opindex mno-specld-anomaly
10530 Don't generate extra code to prevent speculative loads from occurring.
10532 @item -mcsync-anomaly
10533 @opindex mcsync-anomaly
10534 When enabled, the compiler will ensure that the generated code does not
10535 contain CSYNC or SSYNC instructions too soon after conditional branches.
10536 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10538 @item -mno-csync-anomaly
10539 @opindex mno-csync-anomaly
10540 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10541 occurring too soon after a conditional branch.
10545 When enabled, the compiler is free to take advantage of the knowledge that
10546 the entire program fits into the low 64k of memory.
10549 @opindex mno-low-64k
10550 Assume that the program is arbitrarily large. This is the default.
10552 @item -mstack-check-l1
10553 @opindex mstack-check-l1
10554 Do stack checking using information placed into L1 scratchpad memory by the
10557 @item -mid-shared-library
10558 @opindex mid-shared-library
10559 Generate code that supports shared libraries via the library ID method.
10560 This allows for execute in place and shared libraries in an environment
10561 without virtual memory management. This option implies @option{-fPIC}.
10562 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10564 @item -mno-id-shared-library
10565 @opindex mno-id-shared-library
10566 Generate code that doesn't assume ID based shared libraries are being used.
10567 This is the default.
10569 @item -mleaf-id-shared-library
10570 @opindex mleaf-id-shared-library
10571 Generate code that supports shared libraries via the library ID method,
10572 but assumes that this library or executable won't link against any other
10573 ID shared libraries. That allows the compiler to use faster code for jumps
10576 @item -mno-leaf-id-shared-library
10577 @opindex mno-leaf-id-shared-library
10578 Do not assume that the code being compiled won't link against any ID shared
10579 libraries. Slower code will be generated for jump and call insns.
10581 @item -mshared-library-id=n
10582 @opindex mshared-library-id
10583 Specified the identification number of the ID based shared library being
10584 compiled. Specifying a value of 0 will generate more compact code, specifying
10585 other values will force the allocation of that number to the current
10586 library but is no more space or time efficient than omitting this option.
10590 Generate code that allows the data segment to be located in a different
10591 area of memory from the text segment. This allows for execute in place in
10592 an environment without virtual memory management by eliminating relocations
10593 against the text section.
10595 @item -mno-sep-data
10596 @opindex mno-sep-data
10597 Generate code that assumes that the data segment follows the text segment.
10598 This is the default.
10601 @itemx -mno-long-calls
10602 @opindex mlong-calls
10603 @opindex mno-long-calls
10604 Tells the compiler to perform function calls by first loading the
10605 address of the function into a register and then performing a subroutine
10606 call on this register. This switch is needed if the target function
10607 will lie outside of the 24 bit addressing range of the offset based
10608 version of subroutine call instruction.
10610 This feature is not enabled by default. Specifying
10611 @option{-mno-long-calls} will restore the default behavior. Note these
10612 switches have no effect on how the compiler generates code to handle
10613 function calls via function pointers.
10617 Link with the fast floating-point library. This library relaxes some of
10618 the IEEE floating-point standard's rules for checking inputs against
10619 Not-a-Number (NAN), in the interest of performance.
10622 @opindex minline-plt
10623 Enable inlining of PLT entries in function calls to functions that are
10624 not known to bind locally. It has no effect without @option{-mfdpic}.
10627 @opindex mmulticore
10628 Build standalone application for multicore Blackfin processor. Proper
10629 start files and link scripts will be used to support multicore.
10630 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10631 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10632 @option{-mcorea} or @option{-mcoreb}. If it's used without
10633 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10634 programming model is used. In this model, the main function of Core B
10635 should be named as coreb_main. If it's used with @option{-mcorea} or
10636 @option{-mcoreb}, one application per core programming model is used.
10637 If this option is not used, single core application programming
10642 Build standalone application for Core A of BF561 when using
10643 one application per core programming model. Proper start files
10644 and link scripts will be used to support Core A. This option
10645 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10649 Build standalone application for Core B of BF561 when using
10650 one application per core programming model. Proper start files
10651 and link scripts will be used to support Core B. This option
10652 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10653 should be used instead of main. It must be used with
10654 @option{-mmulticore}.
10658 Build standalone application for SDRAM. Proper start files and
10659 link scripts will be used to put the application into SDRAM.
10660 Loader should initialize SDRAM before loading the application
10661 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10665 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10666 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10667 are enabled; for standalone applications the default is off.
10671 @subsection CRIS Options
10672 @cindex CRIS Options
10674 These options are defined specifically for the CRIS ports.
10677 @item -march=@var{architecture-type}
10678 @itemx -mcpu=@var{architecture-type}
10681 Generate code for the specified architecture. The choices for
10682 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10683 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10684 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10687 @item -mtune=@var{architecture-type}
10689 Tune to @var{architecture-type} everything applicable about the generated
10690 code, except for the ABI and the set of available instructions. The
10691 choices for @var{architecture-type} are the same as for
10692 @option{-march=@var{architecture-type}}.
10694 @item -mmax-stack-frame=@var{n}
10695 @opindex mmax-stack-frame
10696 Warn when the stack frame of a function exceeds @var{n} bytes.
10702 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10703 @option{-march=v3} and @option{-march=v8} respectively.
10705 @item -mmul-bug-workaround
10706 @itemx -mno-mul-bug-workaround
10707 @opindex mmul-bug-workaround
10708 @opindex mno-mul-bug-workaround
10709 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10710 models where it applies. This option is active by default.
10714 Enable CRIS-specific verbose debug-related information in the assembly
10715 code. This option also has the effect to turn off the @samp{#NO_APP}
10716 formatted-code indicator to the assembler at the beginning of the
10721 Do not use condition-code results from previous instruction; always emit
10722 compare and test instructions before use of condition codes.
10724 @item -mno-side-effects
10725 @opindex mno-side-effects
10726 Do not emit instructions with side-effects in addressing modes other than
10729 @item -mstack-align
10730 @itemx -mno-stack-align
10731 @itemx -mdata-align
10732 @itemx -mno-data-align
10733 @itemx -mconst-align
10734 @itemx -mno-const-align
10735 @opindex mstack-align
10736 @opindex mno-stack-align
10737 @opindex mdata-align
10738 @opindex mno-data-align
10739 @opindex mconst-align
10740 @opindex mno-const-align
10741 These options (no-options) arranges (eliminate arrangements) for the
10742 stack-frame, individual data and constants to be aligned for the maximum
10743 single data access size for the chosen CPU model. The default is to
10744 arrange for 32-bit alignment. ABI details such as structure layout are
10745 not affected by these options.
10753 Similar to the stack- data- and const-align options above, these options
10754 arrange for stack-frame, writable data and constants to all be 32-bit,
10755 16-bit or 8-bit aligned. The default is 32-bit alignment.
10757 @item -mno-prologue-epilogue
10758 @itemx -mprologue-epilogue
10759 @opindex mno-prologue-epilogue
10760 @opindex mprologue-epilogue
10761 With @option{-mno-prologue-epilogue}, the normal function prologue and
10762 epilogue that sets up the stack-frame are omitted and no return
10763 instructions or return sequences are generated in the code. Use this
10764 option only together with visual inspection of the compiled code: no
10765 warnings or errors are generated when call-saved registers must be saved,
10766 or storage for local variable needs to be allocated.
10770 @opindex mno-gotplt
10772 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10773 instruction sequences that load addresses for functions from the PLT part
10774 of the GOT rather than (traditional on other architectures) calls to the
10775 PLT@. The default is @option{-mgotplt}.
10779 Legacy no-op option only recognized with the cris-axis-elf and
10780 cris-axis-linux-gnu targets.
10784 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10788 This option, recognized for the cris-axis-elf arranges
10789 to link with input-output functions from a simulator library. Code,
10790 initialized data and zero-initialized data are allocated consecutively.
10794 Like @option{-sim}, but pass linker options to locate initialized data at
10795 0x40000000 and zero-initialized data at 0x80000000.
10799 @subsection CRX Options
10800 @cindex CRX Options
10802 These options are defined specifically for the CRX ports.
10808 Enable the use of multiply-accumulate instructions. Disabled by default.
10811 @opindex mpush-args
10812 Push instructions will be used to pass outgoing arguments when functions
10813 are called. Enabled by default.
10816 @node Darwin Options
10817 @subsection Darwin Options
10818 @cindex Darwin options
10820 These options are defined for all architectures running the Darwin operating
10823 FSF GCC on Darwin does not create ``fat'' object files; it will create
10824 an object file for the single architecture that it was built to
10825 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10826 @option{-arch} options are used; it does so by running the compiler or
10827 linker multiple times and joining the results together with
10830 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10831 @samp{i686}) is determined by the flags that specify the ISA
10832 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10833 @option{-force_cpusubtype_ALL} option can be used to override this.
10835 The Darwin tools vary in their behavior when presented with an ISA
10836 mismatch. The assembler, @file{as}, will only permit instructions to
10837 be used that are valid for the subtype of the file it is generating,
10838 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10839 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10840 and print an error if asked to create a shared library with a less
10841 restrictive subtype than its input files (for instance, trying to put
10842 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10843 for executables, @file{ld}, will quietly give the executable the most
10844 restrictive subtype of any of its input files.
10849 Add the framework directory @var{dir} to the head of the list of
10850 directories to be searched for header files. These directories are
10851 interleaved with those specified by @option{-I} options and are
10852 scanned in a left-to-right order.
10854 A framework directory is a directory with frameworks in it. A
10855 framework is a directory with a @samp{"Headers"} and/or
10856 @samp{"PrivateHeaders"} directory contained directly in it that ends
10857 in @samp{".framework"}. The name of a framework is the name of this
10858 directory excluding the @samp{".framework"}. Headers associated with
10859 the framework are found in one of those two directories, with
10860 @samp{"Headers"} being searched first. A subframework is a framework
10861 directory that is in a framework's @samp{"Frameworks"} directory.
10862 Includes of subframework headers can only appear in a header of a
10863 framework that contains the subframework, or in a sibling subframework
10864 header. Two subframeworks are siblings if they occur in the same
10865 framework. A subframework should not have the same name as a
10866 framework, a warning will be issued if this is violated. Currently a
10867 subframework cannot have subframeworks, in the future, the mechanism
10868 may be extended to support this. The standard frameworks can be found
10869 in @samp{"/System/Library/Frameworks"} and
10870 @samp{"/Library/Frameworks"}. An example include looks like
10871 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10872 the name of the framework and header.h is found in the
10873 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10875 @item -iframework@var{dir}
10876 @opindex iframework
10877 Like @option{-F} except the directory is a treated as a system
10878 directory. The main difference between this @option{-iframework} and
10879 @option{-F} is that with @option{-iframework} the compiler does not
10880 warn about constructs contained within header files found via
10881 @var{dir}. This option is valid only for the C family of languages.
10885 Emit debugging information for symbols that are used. For STABS
10886 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10887 This is by default ON@.
10891 Emit debugging information for all symbols and types.
10893 @item -mmacosx-version-min=@var{version}
10894 The earliest version of MacOS X that this executable will run on
10895 is @var{version}. Typical values of @var{version} include @code{10.1},
10896 @code{10.2}, and @code{10.3.9}.
10898 If the compiler was built to use the system's headers by default,
10899 then the default for this option is the system version on which the
10900 compiler is running, otherwise the default is to make choices which
10901 are compatible with as many systems and code bases as possible.
10905 Enable kernel development mode. The @option{-mkernel} option sets
10906 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10907 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10908 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10909 applicable. This mode also sets @option{-mno-altivec},
10910 @option{-msoft-float}, @option{-fno-builtin} and
10911 @option{-mlong-branch} for PowerPC targets.
10913 @item -mone-byte-bool
10914 @opindex mone-byte-bool
10915 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10916 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10917 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10918 option has no effect on x86.
10920 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10921 to generate code that is not binary compatible with code generated
10922 without that switch. Using this switch may require recompiling all
10923 other modules in a program, including system libraries. Use this
10924 switch to conform to a non-default data model.
10926 @item -mfix-and-continue
10927 @itemx -ffix-and-continue
10928 @itemx -findirect-data
10929 @opindex mfix-and-continue
10930 @opindex ffix-and-continue
10931 @opindex findirect-data
10932 Generate code suitable for fast turn around development. Needed to
10933 enable gdb to dynamically load @code{.o} files into already running
10934 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10935 are provided for backwards compatibility.
10939 Loads all members of static archive libraries.
10940 See man ld(1) for more information.
10942 @item -arch_errors_fatal
10943 @opindex arch_errors_fatal
10944 Cause the errors having to do with files that have the wrong architecture
10947 @item -bind_at_load
10948 @opindex bind_at_load
10949 Causes the output file to be marked such that the dynamic linker will
10950 bind all undefined references when the file is loaded or launched.
10954 Produce a Mach-o bundle format file.
10955 See man ld(1) for more information.
10957 @item -bundle_loader @var{executable}
10958 @opindex bundle_loader
10959 This option specifies the @var{executable} that will be loading the build
10960 output file being linked. See man ld(1) for more information.
10963 @opindex dynamiclib
10964 When passed this option, GCC will produce a dynamic library instead of
10965 an executable when linking, using the Darwin @file{libtool} command.
10967 @item -force_cpusubtype_ALL
10968 @opindex force_cpusubtype_ALL
10969 This causes GCC's output file to have the @var{ALL} subtype, instead of
10970 one controlled by the @option{-mcpu} or @option{-march} option.
10972 @item -allowable_client @var{client_name}
10973 @itemx -client_name
10974 @itemx -compatibility_version
10975 @itemx -current_version
10977 @itemx -dependency-file
10979 @itemx -dylinker_install_name
10981 @itemx -exported_symbols_list
10984 @itemx -flat_namespace
10985 @itemx -force_flat_namespace
10986 @itemx -headerpad_max_install_names
10989 @itemx -install_name
10990 @itemx -keep_private_externs
10991 @itemx -multi_module
10992 @itemx -multiply_defined
10993 @itemx -multiply_defined_unused
10996 @itemx -no_dead_strip_inits_and_terms
10997 @itemx -nofixprebinding
10998 @itemx -nomultidefs
11000 @itemx -noseglinkedit
11001 @itemx -pagezero_size
11003 @itemx -prebind_all_twolevel_modules
11004 @itemx -private_bundle
11006 @itemx -read_only_relocs
11008 @itemx -sectobjectsymbols
11012 @itemx -sectobjectsymbols
11015 @itemx -segs_read_only_addr
11017 @itemx -segs_read_write_addr
11018 @itemx -seg_addr_table
11019 @itemx -seg_addr_table_filename
11020 @itemx -seglinkedit
11022 @itemx -segs_read_only_addr
11023 @itemx -segs_read_write_addr
11024 @itemx -single_module
11026 @itemx -sub_library
11028 @itemx -sub_umbrella
11029 @itemx -twolevel_namespace
11032 @itemx -unexported_symbols_list
11033 @itemx -weak_reference_mismatches
11034 @itemx -whatsloaded
11035 @opindex allowable_client
11036 @opindex client_name
11037 @opindex compatibility_version
11038 @opindex current_version
11039 @opindex dead_strip
11040 @opindex dependency-file
11041 @opindex dylib_file
11042 @opindex dylinker_install_name
11044 @opindex exported_symbols_list
11046 @opindex flat_namespace
11047 @opindex force_flat_namespace
11048 @opindex headerpad_max_install_names
11049 @opindex image_base
11051 @opindex install_name
11052 @opindex keep_private_externs
11053 @opindex multi_module
11054 @opindex multiply_defined
11055 @opindex multiply_defined_unused
11056 @opindex noall_load
11057 @opindex no_dead_strip_inits_and_terms
11058 @opindex nofixprebinding
11059 @opindex nomultidefs
11061 @opindex noseglinkedit
11062 @opindex pagezero_size
11064 @opindex prebind_all_twolevel_modules
11065 @opindex private_bundle
11066 @opindex read_only_relocs
11068 @opindex sectobjectsymbols
11071 @opindex sectcreate
11072 @opindex sectobjectsymbols
11075 @opindex segs_read_only_addr
11076 @opindex segs_read_write_addr
11077 @opindex seg_addr_table
11078 @opindex seg_addr_table_filename
11079 @opindex seglinkedit
11081 @opindex segs_read_only_addr
11082 @opindex segs_read_write_addr
11083 @opindex single_module
11085 @opindex sub_library
11086 @opindex sub_umbrella
11087 @opindex twolevel_namespace
11090 @opindex unexported_symbols_list
11091 @opindex weak_reference_mismatches
11092 @opindex whatsloaded
11093 These options are passed to the Darwin linker. The Darwin linker man page
11094 describes them in detail.
11097 @node DEC Alpha Options
11098 @subsection DEC Alpha Options
11100 These @samp{-m} options are defined for the DEC Alpha implementations:
11103 @item -mno-soft-float
11104 @itemx -msoft-float
11105 @opindex mno-soft-float
11106 @opindex msoft-float
11107 Use (do not use) the hardware floating-point instructions for
11108 floating-point operations. When @option{-msoft-float} is specified,
11109 functions in @file{libgcc.a} will be used to perform floating-point
11110 operations. Unless they are replaced by routines that emulate the
11111 floating-point operations, or compiled in such a way as to call such
11112 emulations routines, these routines will issue floating-point
11113 operations. If you are compiling for an Alpha without floating-point
11114 operations, you must ensure that the library is built so as not to call
11117 Note that Alpha implementations without floating-point operations are
11118 required to have floating-point registers.
11121 @itemx -mno-fp-regs
11123 @opindex mno-fp-regs
11124 Generate code that uses (does not use) the floating-point register set.
11125 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
11126 register set is not used, floating point operands are passed in integer
11127 registers as if they were integers and floating-point results are passed
11128 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
11129 so any function with a floating-point argument or return value called by code
11130 compiled with @option{-mno-fp-regs} must also be compiled with that
11133 A typical use of this option is building a kernel that does not use,
11134 and hence need not save and restore, any floating-point registers.
11138 The Alpha architecture implements floating-point hardware optimized for
11139 maximum performance. It is mostly compliant with the IEEE floating
11140 point standard. However, for full compliance, software assistance is
11141 required. This option generates code fully IEEE compliant code
11142 @emph{except} that the @var{inexact-flag} is not maintained (see below).
11143 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
11144 defined during compilation. The resulting code is less efficient but is
11145 able to correctly support denormalized numbers and exceptional IEEE
11146 values such as not-a-number and plus/minus infinity. Other Alpha
11147 compilers call this option @option{-ieee_with_no_inexact}.
11149 @item -mieee-with-inexact
11150 @opindex mieee-with-inexact
11151 This is like @option{-mieee} except the generated code also maintains
11152 the IEEE @var{inexact-flag}. Turning on this option causes the
11153 generated code to implement fully-compliant IEEE math. In addition to
11154 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
11155 macro. On some Alpha implementations the resulting code may execute
11156 significantly slower than the code generated by default. Since there is
11157 very little code that depends on the @var{inexact-flag}, you should
11158 normally not specify this option. Other Alpha compilers call this
11159 option @option{-ieee_with_inexact}.
11161 @item -mfp-trap-mode=@var{trap-mode}
11162 @opindex mfp-trap-mode
11163 This option controls what floating-point related traps are enabled.
11164 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
11165 The trap mode can be set to one of four values:
11169 This is the default (normal) setting. The only traps that are enabled
11170 are the ones that cannot be disabled in software (e.g., division by zero
11174 In addition to the traps enabled by @samp{n}, underflow traps are enabled
11178 Like @samp{u}, but the instructions are marked to be safe for software
11179 completion (see Alpha architecture manual for details).
11182 Like @samp{su}, but inexact traps are enabled as well.
11185 @item -mfp-rounding-mode=@var{rounding-mode}
11186 @opindex mfp-rounding-mode
11187 Selects the IEEE rounding mode. Other Alpha compilers call this option
11188 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
11193 Normal IEEE rounding mode. Floating point numbers are rounded towards
11194 the nearest machine number or towards the even machine number in case
11198 Round towards minus infinity.
11201 Chopped rounding mode. Floating point numbers are rounded towards zero.
11204 Dynamic rounding mode. A field in the floating point control register
11205 (@var{fpcr}, see Alpha architecture reference manual) controls the
11206 rounding mode in effect. The C library initializes this register for
11207 rounding towards plus infinity. Thus, unless your program modifies the
11208 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
11211 @item -mtrap-precision=@var{trap-precision}
11212 @opindex mtrap-precision
11213 In the Alpha architecture, floating point traps are imprecise. This
11214 means without software assistance it is impossible to recover from a
11215 floating trap and program execution normally needs to be terminated.
11216 GCC can generate code that can assist operating system trap handlers
11217 in determining the exact location that caused a floating point trap.
11218 Depending on the requirements of an application, different levels of
11219 precisions can be selected:
11223 Program precision. This option is the default and means a trap handler
11224 can only identify which program caused a floating point exception.
11227 Function precision. The trap handler can determine the function that
11228 caused a floating point exception.
11231 Instruction precision. The trap handler can determine the exact
11232 instruction that caused a floating point exception.
11235 Other Alpha compilers provide the equivalent options called
11236 @option{-scope_safe} and @option{-resumption_safe}.
11238 @item -mieee-conformant
11239 @opindex mieee-conformant
11240 This option marks the generated code as IEEE conformant. You must not
11241 use this option unless you also specify @option{-mtrap-precision=i} and either
11242 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
11243 is to emit the line @samp{.eflag 48} in the function prologue of the
11244 generated assembly file. Under DEC Unix, this has the effect that
11245 IEEE-conformant math library routines will be linked in.
11247 @item -mbuild-constants
11248 @opindex mbuild-constants
11249 Normally GCC examines a 32- or 64-bit integer constant to
11250 see if it can construct it from smaller constants in two or three
11251 instructions. If it cannot, it will output the constant as a literal and
11252 generate code to load it from the data segment at runtime.
11254 Use this option to require GCC to construct @emph{all} integer constants
11255 using code, even if it takes more instructions (the maximum is six).
11257 You would typically use this option to build a shared library dynamic
11258 loader. Itself a shared library, it must relocate itself in memory
11259 before it can find the variables and constants in its own data segment.
11265 Select whether to generate code to be assembled by the vendor-supplied
11266 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11284 Indicate whether GCC should generate code to use the optional BWX,
11285 CIX, FIX and MAX instruction sets. The default is to use the instruction
11286 sets supported by the CPU type specified via @option{-mcpu=} option or that
11287 of the CPU on which GCC was built if none was specified.
11290 @itemx -mfloat-ieee
11291 @opindex mfloat-vax
11292 @opindex mfloat-ieee
11293 Generate code that uses (does not use) VAX F and G floating point
11294 arithmetic instead of IEEE single and double precision.
11296 @item -mexplicit-relocs
11297 @itemx -mno-explicit-relocs
11298 @opindex mexplicit-relocs
11299 @opindex mno-explicit-relocs
11300 Older Alpha assemblers provided no way to generate symbol relocations
11301 except via assembler macros. Use of these macros does not allow
11302 optimal instruction scheduling. GNU binutils as of version 2.12
11303 supports a new syntax that allows the compiler to explicitly mark
11304 which relocations should apply to which instructions. This option
11305 is mostly useful for debugging, as GCC detects the capabilities of
11306 the assembler when it is built and sets the default accordingly.
11309 @itemx -mlarge-data
11310 @opindex msmall-data
11311 @opindex mlarge-data
11312 When @option{-mexplicit-relocs} is in effect, static data is
11313 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11314 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11315 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11316 16-bit relocations off of the @code{$gp} register. This limits the
11317 size of the small data area to 64KB, but allows the variables to be
11318 directly accessed via a single instruction.
11320 The default is @option{-mlarge-data}. With this option the data area
11321 is limited to just below 2GB@. Programs that require more than 2GB of
11322 data must use @code{malloc} or @code{mmap} to allocate the data in the
11323 heap instead of in the program's data segment.
11325 When generating code for shared libraries, @option{-fpic} implies
11326 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11329 @itemx -mlarge-text
11330 @opindex msmall-text
11331 @opindex mlarge-text
11332 When @option{-msmall-text} is used, the compiler assumes that the
11333 code of the entire program (or shared library) fits in 4MB, and is
11334 thus reachable with a branch instruction. When @option{-msmall-data}
11335 is used, the compiler can assume that all local symbols share the
11336 same @code{$gp} value, and thus reduce the number of instructions
11337 required for a function call from 4 to 1.
11339 The default is @option{-mlarge-text}.
11341 @item -mcpu=@var{cpu_type}
11343 Set the instruction set and instruction scheduling parameters for
11344 machine type @var{cpu_type}. You can specify either the @samp{EV}
11345 style name or the corresponding chip number. GCC supports scheduling
11346 parameters for the EV4, EV5 and EV6 family of processors and will
11347 choose the default values for the instruction set from the processor
11348 you specify. If you do not specify a processor type, GCC will default
11349 to the processor on which the compiler was built.
11351 Supported values for @var{cpu_type} are
11357 Schedules as an EV4 and has no instruction set extensions.
11361 Schedules as an EV5 and has no instruction set extensions.
11365 Schedules as an EV5 and supports the BWX extension.
11370 Schedules as an EV5 and supports the BWX and MAX extensions.
11374 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11378 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11381 Native Linux/GNU toolchains also support the value @samp{native},
11382 which selects the best architecture option for the host processor.
11383 @option{-mcpu=native} has no effect if GCC does not recognize
11386 @item -mtune=@var{cpu_type}
11388 Set only the instruction scheduling parameters for machine type
11389 @var{cpu_type}. The instruction set is not changed.
11391 Native Linux/GNU toolchains also support the value @samp{native},
11392 which selects the best architecture option for the host processor.
11393 @option{-mtune=native} has no effect if GCC does not recognize
11396 @item -mmemory-latency=@var{time}
11397 @opindex mmemory-latency
11398 Sets the latency the scheduler should assume for typical memory
11399 references as seen by the application. This number is highly
11400 dependent on the memory access patterns used by the application
11401 and the size of the external cache on the machine.
11403 Valid options for @var{time} are
11407 A decimal number representing clock cycles.
11413 The compiler contains estimates of the number of clock cycles for
11414 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11415 (also called Dcache, Scache, and Bcache), as well as to main memory.
11416 Note that L3 is only valid for EV5.
11421 @node DEC Alpha/VMS Options
11422 @subsection DEC Alpha/VMS Options
11424 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11427 @item -mvms-return-codes
11428 @opindex mvms-return-codes
11429 Return VMS condition codes from main. The default is to return POSIX
11430 style condition (e.g.@: error) codes.
11432 @item -mdebug-main=@var{prefix}
11433 @opindex mdebug-main=@var{prefix}
11434 Flag the first routine whose name starts with @var{prefix} as the main
11435 routine for the debugger.
11439 Default to 64bit memory allocation routines.
11443 @subsection FR30 Options
11444 @cindex FR30 Options
11446 These options are defined specifically for the FR30 port.
11450 @item -msmall-model
11451 @opindex msmall-model
11452 Use the small address space model. This can produce smaller code, but
11453 it does assume that all symbolic values and addresses will fit into a
11458 Assume that run-time support has been provided and so there is no need
11459 to include the simulator library (@file{libsim.a}) on the linker
11465 @subsection FRV Options
11466 @cindex FRV Options
11472 Only use the first 32 general purpose registers.
11477 Use all 64 general purpose registers.
11482 Use only the first 32 floating point registers.
11487 Use all 64 floating point registers
11490 @opindex mhard-float
11492 Use hardware instructions for floating point operations.
11495 @opindex msoft-float
11497 Use library routines for floating point operations.
11502 Dynamically allocate condition code registers.
11507 Do not try to dynamically allocate condition code registers, only
11508 use @code{icc0} and @code{fcc0}.
11513 Change ABI to use double word insns.
11518 Do not use double word instructions.
11523 Use floating point double instructions.
11526 @opindex mno-double
11528 Do not use floating point double instructions.
11533 Use media instructions.
11538 Do not use media instructions.
11543 Use multiply and add/subtract instructions.
11546 @opindex mno-muladd
11548 Do not use multiply and add/subtract instructions.
11553 Select the FDPIC ABI, that uses function descriptors to represent
11554 pointers to functions. Without any PIC/PIE-related options, it
11555 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11556 assumes GOT entries and small data are within a 12-bit range from the
11557 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11558 are computed with 32 bits.
11559 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11562 @opindex minline-plt
11564 Enable inlining of PLT entries in function calls to functions that are
11565 not known to bind locally. It has no effect without @option{-mfdpic}.
11566 It's enabled by default if optimizing for speed and compiling for
11567 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11568 optimization option such as @option{-O3} or above is present in the
11574 Assume a large TLS segment when generating thread-local code.
11579 Do not assume a large TLS segment when generating thread-local code.
11584 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11585 that is known to be in read-only sections. It's enabled by default,
11586 except for @option{-fpic} or @option{-fpie}: even though it may help
11587 make the global offset table smaller, it trades 1 instruction for 4.
11588 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11589 one of which may be shared by multiple symbols, and it avoids the need
11590 for a GOT entry for the referenced symbol, so it's more likely to be a
11591 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11593 @item -multilib-library-pic
11594 @opindex multilib-library-pic
11596 Link with the (library, not FD) pic libraries. It's implied by
11597 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11598 @option{-fpic} without @option{-mfdpic}. You should never have to use
11602 @opindex mlinked-fp
11604 Follow the EABI requirement of always creating a frame pointer whenever
11605 a stack frame is allocated. This option is enabled by default and can
11606 be disabled with @option{-mno-linked-fp}.
11609 @opindex mlong-calls
11611 Use indirect addressing to call functions outside the current
11612 compilation unit. This allows the functions to be placed anywhere
11613 within the 32-bit address space.
11615 @item -malign-labels
11616 @opindex malign-labels
11618 Try to align labels to an 8-byte boundary by inserting nops into the
11619 previous packet. This option only has an effect when VLIW packing
11620 is enabled. It doesn't create new packets; it merely adds nops to
11623 @item -mlibrary-pic
11624 @opindex mlibrary-pic
11626 Generate position-independent EABI code.
11631 Use only the first four media accumulator registers.
11636 Use all eight media accumulator registers.
11641 Pack VLIW instructions.
11646 Do not pack VLIW instructions.
11649 @opindex mno-eflags
11651 Do not mark ABI switches in e_flags.
11654 @opindex mcond-move
11656 Enable the use of conditional-move instructions (default).
11658 This switch is mainly for debugging the compiler and will likely be removed
11659 in a future version.
11661 @item -mno-cond-move
11662 @opindex mno-cond-move
11664 Disable the use of conditional-move instructions.
11666 This switch is mainly for debugging the compiler and will likely be removed
11667 in a future version.
11672 Enable the use of conditional set instructions (default).
11674 This switch is mainly for debugging the compiler and will likely be removed
11675 in a future version.
11680 Disable the use of conditional set instructions.
11682 This switch is mainly for debugging the compiler and will likely be removed
11683 in a future version.
11686 @opindex mcond-exec
11688 Enable the use of conditional execution (default).
11690 This switch is mainly for debugging the compiler and will likely be removed
11691 in a future version.
11693 @item -mno-cond-exec
11694 @opindex mno-cond-exec
11696 Disable the use of conditional execution.
11698 This switch is mainly for debugging the compiler and will likely be removed
11699 in a future version.
11701 @item -mvliw-branch
11702 @opindex mvliw-branch
11704 Run a pass to pack branches into VLIW instructions (default).
11706 This switch is mainly for debugging the compiler and will likely be removed
11707 in a future version.
11709 @item -mno-vliw-branch
11710 @opindex mno-vliw-branch
11712 Do not run a pass to pack branches into VLIW instructions.
11714 This switch is mainly for debugging the compiler and will likely be removed
11715 in a future version.
11717 @item -mmulti-cond-exec
11718 @opindex mmulti-cond-exec
11720 Enable optimization of @code{&&} and @code{||} in conditional execution
11723 This switch is mainly for debugging the compiler and will likely be removed
11724 in a future version.
11726 @item -mno-multi-cond-exec
11727 @opindex mno-multi-cond-exec
11729 Disable optimization of @code{&&} and @code{||} in conditional execution.
11731 This switch is mainly for debugging the compiler and will likely be removed
11732 in a future version.
11734 @item -mnested-cond-exec
11735 @opindex mnested-cond-exec
11737 Enable nested conditional execution optimizations (default).
11739 This switch is mainly for debugging the compiler and will likely be removed
11740 in a future version.
11742 @item -mno-nested-cond-exec
11743 @opindex mno-nested-cond-exec
11745 Disable nested conditional execution optimizations.
11747 This switch is mainly for debugging the compiler and will likely be removed
11748 in a future version.
11750 @item -moptimize-membar
11751 @opindex moptimize-membar
11753 This switch removes redundant @code{membar} instructions from the
11754 compiler generated code. It is enabled by default.
11756 @item -mno-optimize-membar
11757 @opindex mno-optimize-membar
11759 This switch disables the automatic removal of redundant @code{membar}
11760 instructions from the generated code.
11762 @item -mtomcat-stats
11763 @opindex mtomcat-stats
11765 Cause gas to print out tomcat statistics.
11767 @item -mcpu=@var{cpu}
11770 Select the processor type for which to generate code. Possible values are
11771 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11772 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11776 @node GNU/Linux Options
11777 @subsection GNU/Linux Options
11779 These @samp{-m} options are defined for GNU/Linux targets:
11784 Use the GNU C library. This is the default except
11785 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
11789 Use uClibc C library. This is the default on
11790 @samp{*-*-linux-*uclibc*} targets.
11794 Use Bionic C library. This is the default on
11795 @samp{*-*-linux-*android*} targets.
11799 Compile code compatible with Android platform. This is the default on
11800 @samp{*-*-linux-*android*} targets.
11802 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
11803 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
11804 this option makes the GCC driver pass Android-specific options to the linker.
11805 Finally, this option causes the preprocessor macro @code{__ANDROID__}
11808 @item -tno-android-cc
11809 @opindex tno-android-cc
11810 Disable compilation effects of @option{-mandroid}, i.e., do not enable
11811 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
11812 @option{-fno-rtti} by default.
11814 @item -tno-android-ld
11815 @opindex tno-android-ld
11816 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
11817 linking options to the linker.
11821 @node H8/300 Options
11822 @subsection H8/300 Options
11824 These @samp{-m} options are defined for the H8/300 implementations:
11829 Shorten some address references at link time, when possible; uses the
11830 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11831 ld, Using ld}, for a fuller description.
11835 Generate code for the H8/300H@.
11839 Generate code for the H8S@.
11843 Generate code for the H8S and H8/300H in the normal mode. This switch
11844 must be used either with @option{-mh} or @option{-ms}.
11848 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11852 Make @code{int} data 32 bits by default.
11855 @opindex malign-300
11856 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11857 The default for the H8/300H and H8S is to align longs and floats on 4
11859 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11860 This option has no effect on the H8/300.
11864 @subsection HPPA Options
11865 @cindex HPPA Options
11867 These @samp{-m} options are defined for the HPPA family of computers:
11870 @item -march=@var{architecture-type}
11872 Generate code for the specified architecture. The choices for
11873 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11874 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11875 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11876 architecture option for your machine. Code compiled for lower numbered
11877 architectures will run on higher numbered architectures, but not the
11880 @item -mpa-risc-1-0
11881 @itemx -mpa-risc-1-1
11882 @itemx -mpa-risc-2-0
11883 @opindex mpa-risc-1-0
11884 @opindex mpa-risc-1-1
11885 @opindex mpa-risc-2-0
11886 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11889 @opindex mbig-switch
11890 Generate code suitable for big switch tables. Use this option only if
11891 the assembler/linker complain about out of range branches within a switch
11894 @item -mjump-in-delay
11895 @opindex mjump-in-delay
11896 Fill delay slots of function calls with unconditional jump instructions
11897 by modifying the return pointer for the function call to be the target
11898 of the conditional jump.
11900 @item -mdisable-fpregs
11901 @opindex mdisable-fpregs
11902 Prevent floating point registers from being used in any manner. This is
11903 necessary for compiling kernels which perform lazy context switching of
11904 floating point registers. If you use this option and attempt to perform
11905 floating point operations, the compiler will abort.
11907 @item -mdisable-indexing
11908 @opindex mdisable-indexing
11909 Prevent the compiler from using indexing address modes. This avoids some
11910 rather obscure problems when compiling MIG generated code under MACH@.
11912 @item -mno-space-regs
11913 @opindex mno-space-regs
11914 Generate code that assumes the target has no space registers. This allows
11915 GCC to generate faster indirect calls and use unscaled index address modes.
11917 Such code is suitable for level 0 PA systems and kernels.
11919 @item -mfast-indirect-calls
11920 @opindex mfast-indirect-calls
11921 Generate code that assumes calls never cross space boundaries. This
11922 allows GCC to emit code which performs faster indirect calls.
11924 This option will not work in the presence of shared libraries or nested
11927 @item -mfixed-range=@var{register-range}
11928 @opindex mfixed-range
11929 Generate code treating the given register range as fixed registers.
11930 A fixed register is one that the register allocator can not use. This is
11931 useful when compiling kernel code. A register range is specified as
11932 two registers separated by a dash. Multiple register ranges can be
11933 specified separated by a comma.
11935 @item -mlong-load-store
11936 @opindex mlong-load-store
11937 Generate 3-instruction load and store sequences as sometimes required by
11938 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11941 @item -mportable-runtime
11942 @opindex mportable-runtime
11943 Use the portable calling conventions proposed by HP for ELF systems.
11947 Enable the use of assembler directives only GAS understands.
11949 @item -mschedule=@var{cpu-type}
11951 Schedule code according to the constraints for the machine type
11952 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11953 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11954 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11955 proper scheduling option for your machine. The default scheduling is
11959 @opindex mlinker-opt
11960 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11961 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11962 linkers in which they give bogus error messages when linking some programs.
11965 @opindex msoft-float
11966 Generate output containing library calls for floating point.
11967 @strong{Warning:} the requisite libraries are not available for all HPPA
11968 targets. Normally the facilities of the machine's usual C compiler are
11969 used, but this cannot be done directly in cross-compilation. You must make
11970 your own arrangements to provide suitable library functions for
11973 @option{-msoft-float} changes the calling convention in the output file;
11974 therefore, it is only useful if you compile @emph{all} of a program with
11975 this option. In particular, you need to compile @file{libgcc.a}, the
11976 library that comes with GCC, with @option{-msoft-float} in order for
11981 Generate the predefine, @code{_SIO}, for server IO@. The default is
11982 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11983 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11984 options are available under HP-UX and HI-UX@.
11988 Use GNU ld specific options. This passes @option{-shared} to ld when
11989 building a shared library. It is the default when GCC is configured,
11990 explicitly or implicitly, with the GNU linker. This option does not
11991 have any affect on which ld is called, it only changes what parameters
11992 are passed to that ld. The ld that is called is determined by the
11993 @option{--with-ld} configure option, GCC's program search path, and
11994 finally by the user's @env{PATH}. The linker used by GCC can be printed
11995 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11996 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12000 Use HP ld specific options. This passes @option{-b} to ld when building
12001 a shared library and passes @option{+Accept TypeMismatch} to ld on all
12002 links. It is the default when GCC is configured, explicitly or
12003 implicitly, with the HP linker. This option does not have any affect on
12004 which ld is called, it only changes what parameters are passed to that
12005 ld. The ld that is called is determined by the @option{--with-ld}
12006 configure option, GCC's program search path, and finally by the user's
12007 @env{PATH}. The linker used by GCC can be printed using @samp{which
12008 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
12009 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12012 @opindex mno-long-calls
12013 Generate code that uses long call sequences. This ensures that a call
12014 is always able to reach linker generated stubs. The default is to generate
12015 long calls only when the distance from the call site to the beginning
12016 of the function or translation unit, as the case may be, exceeds a
12017 predefined limit set by the branch type being used. The limits for
12018 normal calls are 7,600,000 and 240,000 bytes, respectively for the
12019 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
12022 Distances are measured from the beginning of functions when using the
12023 @option{-ffunction-sections} option, or when using the @option{-mgas}
12024 and @option{-mno-portable-runtime} options together under HP-UX with
12027 It is normally not desirable to use this option as it will degrade
12028 performance. However, it may be useful in large applications,
12029 particularly when partial linking is used to build the application.
12031 The types of long calls used depends on the capabilities of the
12032 assembler and linker, and the type of code being generated. The
12033 impact on systems that support long absolute calls, and long pic
12034 symbol-difference or pc-relative calls should be relatively small.
12035 However, an indirect call is used on 32-bit ELF systems in pic code
12036 and it is quite long.
12038 @item -munix=@var{unix-std}
12040 Generate compiler predefines and select a startfile for the specified
12041 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
12042 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
12043 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
12044 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
12045 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
12048 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
12049 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
12050 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
12051 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
12052 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
12053 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
12055 It is @emph{important} to note that this option changes the interfaces
12056 for various library routines. It also affects the operational behavior
12057 of the C library. Thus, @emph{extreme} care is needed in using this
12060 Library code that is intended to operate with more than one UNIX
12061 standard must test, set and restore the variable @var{__xpg4_extended_mask}
12062 as appropriate. Most GNU software doesn't provide this capability.
12066 Suppress the generation of link options to search libdld.sl when the
12067 @option{-static} option is specified on HP-UX 10 and later.
12071 The HP-UX implementation of setlocale in libc has a dependency on
12072 libdld.sl. There isn't an archive version of libdld.sl. Thus,
12073 when the @option{-static} option is specified, special link options
12074 are needed to resolve this dependency.
12076 On HP-UX 10 and later, the GCC driver adds the necessary options to
12077 link with libdld.sl when the @option{-static} option is specified.
12078 This causes the resulting binary to be dynamic. On the 64-bit port,
12079 the linkers generate dynamic binaries by default in any case. The
12080 @option{-nolibdld} option can be used to prevent the GCC driver from
12081 adding these link options.
12085 Add support for multithreading with the @dfn{dce thread} library
12086 under HP-UX@. This option sets flags for both the preprocessor and
12090 @node i386 and x86-64 Options
12091 @subsection Intel 386 and AMD x86-64 Options
12092 @cindex i386 Options
12093 @cindex x86-64 Options
12094 @cindex Intel 386 Options
12095 @cindex AMD x86-64 Options
12097 These @samp{-m} options are defined for the i386 and x86-64 family of
12101 @item -mtune=@var{cpu-type}
12103 Tune to @var{cpu-type} everything applicable about the generated code, except
12104 for the ABI and the set of available instructions. The choices for
12105 @var{cpu-type} are:
12108 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
12109 If you know the CPU on which your code will run, then you should use
12110 the corresponding @option{-mtune} option instead of
12111 @option{-mtune=generic}. But, if you do not know exactly what CPU users
12112 of your application will have, then you should use this option.
12114 As new processors are deployed in the marketplace, the behavior of this
12115 option will change. Therefore, if you upgrade to a newer version of
12116 GCC, the code generated option will change to reflect the processors
12117 that were most common when that version of GCC was released.
12119 There is no @option{-march=generic} option because @option{-march}
12120 indicates the instruction set the compiler can use, and there is no
12121 generic instruction set applicable to all processors. In contrast,
12122 @option{-mtune} indicates the processor (or, in this case, collection of
12123 processors) for which the code is optimized.
12125 This selects the CPU to tune for at compilation time by determining
12126 the processor type of the compiling machine. Using @option{-mtune=native}
12127 will produce code optimized for the local machine under the constraints
12128 of the selected instruction set. Using @option{-march=native} will
12129 enable all instruction subsets supported by the local machine (hence
12130 the result might not run on different machines).
12132 Original Intel's i386 CPU@.
12134 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
12135 @item i586, pentium
12136 Intel Pentium CPU with no MMX support.
12138 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
12140 Intel PentiumPro CPU@.
12142 Same as @code{generic}, but when used as @code{march} option, PentiumPro
12143 instruction set will be used, so the code will run on all i686 family chips.
12145 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
12146 @item pentium3, pentium3m
12147 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
12150 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
12151 support. Used by Centrino notebooks.
12152 @item pentium4, pentium4m
12153 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
12155 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
12158 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
12159 SSE2 and SSE3 instruction set support.
12161 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12162 instruction set support.
12164 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
12165 and SSE4.2 instruction set support.
12167 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12168 SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support.
12170 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12171 instruction set support.
12173 AMD K6 CPU with MMX instruction set support.
12175 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
12176 @item athlon, athlon-tbird
12177 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
12179 @item athlon-4, athlon-xp, athlon-mp
12180 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
12181 instruction set support.
12182 @item k8, opteron, athlon64, athlon-fx
12183 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
12184 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
12185 @item k8-sse3, opteron-sse3, athlon64-sse3
12186 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
12187 @item amdfam10, barcelona
12188 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
12189 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
12190 instruction set extensions.)
12192 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
12195 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
12196 instruction set support.
12198 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
12199 implemented for this chip.)
12201 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
12202 implemented for this chip.)
12204 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
12207 While picking a specific @var{cpu-type} will schedule things appropriately
12208 for that particular chip, the compiler will not generate any code that
12209 does not run on the i386 without the @option{-march=@var{cpu-type}} option
12212 @item -march=@var{cpu-type}
12214 Generate instructions for the machine type @var{cpu-type}. The choices
12215 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
12216 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
12218 @item -mcpu=@var{cpu-type}
12220 A deprecated synonym for @option{-mtune}.
12222 @item -mfpmath=@var{unit}
12224 Generate floating point arithmetics for selected unit @var{unit}. The choices
12225 for @var{unit} are:
12229 Use the standard 387 floating point coprocessor present majority of chips and
12230 emulated otherwise. Code compiled with this option will run almost everywhere.
12231 The temporary results are computed in 80bit precision instead of precision
12232 specified by the type resulting in slightly different results compared to most
12233 of other chips. See @option{-ffloat-store} for more detailed description.
12235 This is the default choice for i386 compiler.
12238 Use scalar floating point instructions present in the SSE instruction set.
12239 This instruction set is supported by Pentium3 and newer chips, in the AMD line
12240 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
12241 instruction set supports only single precision arithmetics, thus the double and
12242 extended precision arithmetics is still done using 387. Later version, present
12243 only in Pentium4 and the future AMD x86-64 chips supports double precision
12246 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
12247 or @option{-msse2} switches to enable SSE extensions and make this option
12248 effective. For the x86-64 compiler, these extensions are enabled by default.
12250 The resulting code should be considerably faster in the majority of cases and avoid
12251 the numerical instability problems of 387 code, but may break some existing
12252 code that expects temporaries to be 80bit.
12254 This is the default choice for the x86-64 compiler.
12259 Attempt to utilize both instruction sets at once. This effectively double the
12260 amount of available registers and on chips with separate execution units for
12261 387 and SSE the execution resources too. Use this option with care, as it is
12262 still experimental, because the GCC register allocator does not model separate
12263 functional units well resulting in instable performance.
12266 @item -masm=@var{dialect}
12267 @opindex masm=@var{dialect}
12268 Output asm instructions using selected @var{dialect}. Supported
12269 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12270 not support @samp{intel}.
12273 @itemx -mno-ieee-fp
12275 @opindex mno-ieee-fp
12276 Control whether or not the compiler uses IEEE floating point
12277 comparisons. These handle correctly the case where the result of a
12278 comparison is unordered.
12281 @opindex msoft-float
12282 Generate output containing library calls for floating point.
12283 @strong{Warning:} the requisite libraries are not part of GCC@.
12284 Normally the facilities of the machine's usual C compiler are used, but
12285 this can't be done directly in cross-compilation. You must make your
12286 own arrangements to provide suitable library functions for
12289 On machines where a function returns floating point results in the 80387
12290 register stack, some floating point opcodes may be emitted even if
12291 @option{-msoft-float} is used.
12293 @item -mno-fp-ret-in-387
12294 @opindex mno-fp-ret-in-387
12295 Do not use the FPU registers for return values of functions.
12297 The usual calling convention has functions return values of types
12298 @code{float} and @code{double} in an FPU register, even if there
12299 is no FPU@. The idea is that the operating system should emulate
12302 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12303 in ordinary CPU registers instead.
12305 @item -mno-fancy-math-387
12306 @opindex mno-fancy-math-387
12307 Some 387 emulators do not support the @code{sin}, @code{cos} and
12308 @code{sqrt} instructions for the 387. Specify this option to avoid
12309 generating those instructions. This option is the default on FreeBSD,
12310 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12311 indicates that the target CPU will always have an FPU and so the
12312 instruction will not need emulation. As of revision 2.6.1, these
12313 instructions are not generated unless you also use the
12314 @option{-funsafe-math-optimizations} switch.
12316 @item -malign-double
12317 @itemx -mno-align-double
12318 @opindex malign-double
12319 @opindex mno-align-double
12320 Control whether GCC aligns @code{double}, @code{long double}, and
12321 @code{long long} variables on a two word boundary or a one word
12322 boundary. Aligning @code{double} variables on a two word boundary will
12323 produce code that runs somewhat faster on a @samp{Pentium} at the
12324 expense of more memory.
12326 On x86-64, @option{-malign-double} is enabled by default.
12328 @strong{Warning:} if you use the @option{-malign-double} switch,
12329 structures containing the above types will be aligned differently than
12330 the published application binary interface specifications for the 386
12331 and will not be binary compatible with structures in code compiled
12332 without that switch.
12334 @item -m96bit-long-double
12335 @itemx -m128bit-long-double
12336 @opindex m96bit-long-double
12337 @opindex m128bit-long-double
12338 These switches control the size of @code{long double} type. The i386
12339 application binary interface specifies the size to be 96 bits,
12340 so @option{-m96bit-long-double} is the default in 32 bit mode.
12342 Modern architectures (Pentium and newer) would prefer @code{long double}
12343 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12344 conforming to the ABI, this would not be possible. So specifying a
12345 @option{-m128bit-long-double} will align @code{long double}
12346 to a 16 byte boundary by padding the @code{long double} with an additional
12349 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12350 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12352 Notice that neither of these options enable any extra precision over the x87
12353 standard of 80 bits for a @code{long double}.
12355 @strong{Warning:} if you override the default value for your target ABI, the
12356 structures and arrays containing @code{long double} variables will change
12357 their size as well as function calling convention for function taking
12358 @code{long double} will be modified. Hence they will not be binary
12359 compatible with arrays or structures in code compiled without that switch.
12361 @item -mlarge-data-threshold=@var{number}
12362 @opindex mlarge-data-threshold=@var{number}
12363 When @option{-mcmodel=medium} is specified, the data greater than
12364 @var{threshold} are placed in large data section. This value must be the
12365 same across all object linked into the binary and defaults to 65535.
12369 Use a different function-calling convention, in which functions that
12370 take a fixed number of arguments return with the @code{ret} @var{num}
12371 instruction, which pops their arguments while returning. This saves one
12372 instruction in the caller since there is no need to pop the arguments
12375 You can specify that an individual function is called with this calling
12376 sequence with the function attribute @samp{stdcall}. You can also
12377 override the @option{-mrtd} option by using the function attribute
12378 @samp{cdecl}. @xref{Function Attributes}.
12380 @strong{Warning:} this calling convention is incompatible with the one
12381 normally used on Unix, so you cannot use it if you need to call
12382 libraries compiled with the Unix compiler.
12384 Also, you must provide function prototypes for all functions that
12385 take variable numbers of arguments (including @code{printf});
12386 otherwise incorrect code will be generated for calls to those
12389 In addition, seriously incorrect code will result if you call a
12390 function with too many arguments. (Normally, extra arguments are
12391 harmlessly ignored.)
12393 @item -mregparm=@var{num}
12395 Control how many registers are used to pass integer arguments. By
12396 default, no registers are used to pass arguments, and at most 3
12397 registers can be used. You can control this behavior for a specific
12398 function by using the function attribute @samp{regparm}.
12399 @xref{Function Attributes}.
12401 @strong{Warning:} if you use this switch, and
12402 @var{num} is nonzero, then you must build all modules with the same
12403 value, including any libraries. This includes the system libraries and
12407 @opindex msseregparm
12408 Use SSE register passing conventions for float and double arguments
12409 and return values. You can control this behavior for a specific
12410 function by using the function attribute @samp{sseregparm}.
12411 @xref{Function Attributes}.
12413 @strong{Warning:} if you use this switch then you must build all
12414 modules with the same value, including any libraries. This includes
12415 the system libraries and startup modules.
12417 @item -mvect8-ret-in-mem
12418 @opindex mvect8-ret-in-mem
12419 Return 8-byte vectors in memory instead of MMX registers. This is the
12420 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
12421 Studio compilers until version 12. Later compiler versions (starting
12422 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
12423 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
12424 you need to remain compatible with existing code produced by those
12425 previous compiler versions or older versions of GCC.
12434 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
12435 is specified, the significands of results of floating-point operations are
12436 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12437 significands of results of floating-point operations to 53 bits (double
12438 precision) and @option{-mpc80} rounds the significands of results of
12439 floating-point operations to 64 bits (extended double precision), which is
12440 the default. When this option is used, floating-point operations in higher
12441 precisions are not available to the programmer without setting the FPU
12442 control word explicitly.
12444 Setting the rounding of floating-point operations to less than the default
12445 80 bits can speed some programs by 2% or more. Note that some mathematical
12446 libraries assume that extended precision (80 bit) floating-point operations
12447 are enabled by default; routines in such libraries could suffer significant
12448 loss of accuracy, typically through so-called "catastrophic cancellation",
12449 when this option is used to set the precision to less than extended precision.
12451 @item -mstackrealign
12452 @opindex mstackrealign
12453 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12454 option will generate an alternate prologue and epilogue that realigns the
12455 runtime stack if necessary. This supports mixing legacy codes that keep
12456 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12457 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12458 applicable to individual functions.
12460 @item -mpreferred-stack-boundary=@var{num}
12461 @opindex mpreferred-stack-boundary
12462 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12463 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12464 the default is 4 (16 bytes or 128 bits).
12466 @item -mincoming-stack-boundary=@var{num}
12467 @opindex mincoming-stack-boundary
12468 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12469 boundary. If @option{-mincoming-stack-boundary} is not specified,
12470 the one specified by @option{-mpreferred-stack-boundary} will be used.
12472 On Pentium and PentiumPro, @code{double} and @code{long double} values
12473 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12474 suffer significant run time performance penalties. On Pentium III, the
12475 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12476 properly if it is not 16 byte aligned.
12478 To ensure proper alignment of this values on the stack, the stack boundary
12479 must be as aligned as that required by any value stored on the stack.
12480 Further, every function must be generated such that it keeps the stack
12481 aligned. Thus calling a function compiled with a higher preferred
12482 stack boundary from a function compiled with a lower preferred stack
12483 boundary will most likely misalign the stack. It is recommended that
12484 libraries that use callbacks always use the default setting.
12486 This extra alignment does consume extra stack space, and generally
12487 increases code size. Code that is sensitive to stack space usage, such
12488 as embedded systems and operating system kernels, may want to reduce the
12489 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12516 @itemx -mno-fsgsbase
12546 These switches enable or disable the use of instructions in the MMX,
12547 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, FSGSBASE, RDRND,
12548 F16C, SSE4A, FMA4, XOP, LWP, ABM, BMI, or 3DNow!@: extended instruction sets.
12549 These extensions are also available as built-in functions: see
12550 @ref{X86 Built-in Functions}, for details of the functions enabled and
12551 disabled by these switches.
12553 To have SSE/SSE2 instructions generated automatically from floating-point
12554 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12556 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12557 generates new AVX instructions or AVX equivalence for all SSEx instructions
12560 These options will enable GCC to use these extended instructions in
12561 generated code, even without @option{-mfpmath=sse}. Applications which
12562 perform runtime CPU detection must compile separate files for each
12563 supported architecture, using the appropriate flags. In particular,
12564 the file containing the CPU detection code should be compiled without
12568 @itemx -mno-fused-madd
12569 @opindex mfused-madd
12570 @opindex mno-fused-madd
12571 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12572 instructions. The default is to use these instructions.
12576 This option instructs GCC to emit a @code{cld} instruction in the prologue
12577 of functions that use string instructions. String instructions depend on
12578 the DF flag to select between autoincrement or autodecrement mode. While the
12579 ABI specifies the DF flag to be cleared on function entry, some operating
12580 systems violate this specification by not clearing the DF flag in their
12581 exception dispatchers. The exception handler can be invoked with the DF flag
12582 set which leads to wrong direction mode, when string instructions are used.
12583 This option can be enabled by default on 32-bit x86 targets by configuring
12584 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12585 instructions can be suppressed with the @option{-mno-cld} compiler option
12589 @opindex mvzeroupper
12590 This option instructs GCC to emit a @code{vzeroupper} instruction
12591 before a transfer of control flow out of the function to minimize
12592 AVX to SSE transition penalty as well as remove unnecessary zeroupper
12597 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12598 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12599 data types. This is useful for high resolution counters that could be updated
12600 by multiple processors (or cores). This instruction is generated as part of
12601 atomic built-in functions: see @ref{Atomic Builtins} for details.
12605 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12606 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12607 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12608 SAHF are load and store instructions, respectively, for certain status flags.
12609 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12610 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12614 This option will enable GCC to use movbe instruction to implement
12615 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12619 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12620 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12621 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12625 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12626 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12627 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12628 variants) for single precision floating point arguments. These instructions
12629 are generated only when @option{-funsafe-math-optimizations} is enabled
12630 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12631 Note that while the throughput of the sequence is higher than the throughput
12632 of the non-reciprocal instruction, the precision of the sequence can be
12633 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12635 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12636 already with @option{-ffast-math} (or the above option combination), and
12637 doesn't need @option{-mrecip}.
12639 @item -mveclibabi=@var{type}
12640 @opindex mveclibabi
12641 Specifies the ABI type to use for vectorizing intrinsics using an
12642 external library. Supported types are @code{svml} for the Intel short
12643 vector math library and @code{acml} for the AMD math core library style
12644 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12645 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12646 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12647 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12648 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12649 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12650 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12651 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12652 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12653 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12654 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12655 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12656 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12657 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12658 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12659 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12660 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12661 compatible library will have to be specified at link time.
12663 @item -mabi=@var{name}
12665 Generate code for the specified calling convention. Permissible values
12666 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12667 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12668 ABI when targeting Windows. On all other systems, the default is the
12669 SYSV ABI. You can control this behavior for a specific function by
12670 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12671 @xref{Function Attributes}.
12674 @itemx -mno-push-args
12675 @opindex mpush-args
12676 @opindex mno-push-args
12677 Use PUSH operations to store outgoing parameters. This method is shorter
12678 and usually equally fast as method using SUB/MOV operations and is enabled
12679 by default. In some cases disabling it may improve performance because of
12680 improved scheduling and reduced dependencies.
12682 @item -maccumulate-outgoing-args
12683 @opindex maccumulate-outgoing-args
12684 If enabled, the maximum amount of space required for outgoing arguments will be
12685 computed in the function prologue. This is faster on most modern CPUs
12686 because of reduced dependencies, improved scheduling and reduced stack usage
12687 when preferred stack boundary is not equal to 2. The drawback is a notable
12688 increase in code size. This switch implies @option{-mno-push-args}.
12692 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12693 on thread-safe exception handling must compile and link all code with the
12694 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12695 @option{-D_MT}; when linking, it links in a special thread helper library
12696 @option{-lmingwthrd} which cleans up per thread exception handling data.
12698 @item -mno-align-stringops
12699 @opindex mno-align-stringops
12700 Do not align destination of inlined string operations. This switch reduces
12701 code size and improves performance in case the destination is already aligned,
12702 but GCC doesn't know about it.
12704 @item -minline-all-stringops
12705 @opindex minline-all-stringops
12706 By default GCC inlines string operations only when destination is known to be
12707 aligned at least to 4 byte boundary. This enables more inlining, increase code
12708 size, but may improve performance of code that depends on fast memcpy, strlen
12709 and memset for short lengths.
12711 @item -minline-stringops-dynamically
12712 @opindex minline-stringops-dynamically
12713 For string operation of unknown size, inline runtime checks so for small
12714 blocks inline code is used, while for large blocks library call is used.
12716 @item -mstringop-strategy=@var{alg}
12717 @opindex mstringop-strategy=@var{alg}
12718 Overwrite internal decision heuristic about particular algorithm to inline
12719 string operation with. The allowed values are @code{rep_byte},
12720 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12721 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12722 expanding inline loop, @code{libcall} for always expanding library call.
12724 @item -momit-leaf-frame-pointer
12725 @opindex momit-leaf-frame-pointer
12726 Don't keep the frame pointer in a register for leaf functions. This
12727 avoids the instructions to save, set up and restore frame pointers and
12728 makes an extra register available in leaf functions. The option
12729 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12730 which might make debugging harder.
12732 @item -mtls-direct-seg-refs
12733 @itemx -mno-tls-direct-seg-refs
12734 @opindex mtls-direct-seg-refs
12735 Controls whether TLS variables may be accessed with offsets from the
12736 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12737 or whether the thread base pointer must be added. Whether or not this
12738 is legal depends on the operating system, and whether it maps the
12739 segment to cover the entire TLS area.
12741 For systems that use GNU libc, the default is on.
12744 @itemx -mno-sse2avx
12746 Specify that the assembler should encode SSE instructions with VEX
12747 prefix. The option @option{-mavx} turns this on by default.
12752 If profiling is active @option{-pg} put the profiling
12753 counter call before prologue.
12754 Note: On x86 architectures the attribute @code{ms_hook_prologue}
12755 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
12758 @itemx -mno-8bit-idiv
12760 On some processors, like Intel Atom, 8bit unsigned integer divide is
12761 much faster than 32bit/64bit integer divide. This option will generate a
12762 runt-time check. If both dividend and divisor are within range of 0
12763 to 255, 8bit unsigned integer divide will be used instead of
12764 32bit/64bit integer divide.
12768 These @samp{-m} switches are supported in addition to the above
12769 on AMD x86-64 processors in 64-bit environments.
12776 Generate code for a 32-bit or 64-bit environment.
12777 The 32-bit environment sets int, long and pointer to 32 bits and
12778 generates code that runs on any i386 system.
12779 The 64-bit environment sets int to 32 bits and long and pointer
12780 to 64 bits and generates code for AMD's x86-64 architecture. For
12781 darwin only the -m64 option turns off the @option{-fno-pic} and
12782 @option{-mdynamic-no-pic} options.
12784 @item -mno-red-zone
12785 @opindex mno-red-zone
12786 Do not use a so called red zone for x86-64 code. The red zone is mandated
12787 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12788 stack pointer that will not be modified by signal or interrupt handlers
12789 and therefore can be used for temporary data without adjusting the stack
12790 pointer. The flag @option{-mno-red-zone} disables this red zone.
12792 @item -mcmodel=small
12793 @opindex mcmodel=small
12794 Generate code for the small code model: the program and its symbols must
12795 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12796 Programs can be statically or dynamically linked. This is the default
12799 @item -mcmodel=kernel
12800 @opindex mcmodel=kernel
12801 Generate code for the kernel code model. The kernel runs in the
12802 negative 2 GB of the address space.
12803 This model has to be used for Linux kernel code.
12805 @item -mcmodel=medium
12806 @opindex mcmodel=medium
12807 Generate code for the medium model: The program is linked in the lower 2
12808 GB of the address space. Small symbols are also placed there. Symbols
12809 with sizes larger than @option{-mlarge-data-threshold} are put into
12810 large data or bss sections and can be located above 2GB. Programs can
12811 be statically or dynamically linked.
12813 @item -mcmodel=large
12814 @opindex mcmodel=large
12815 Generate code for the large model: This model makes no assumptions
12816 about addresses and sizes of sections.
12819 @node IA-64 Options
12820 @subsection IA-64 Options
12821 @cindex IA-64 Options
12823 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12827 @opindex mbig-endian
12828 Generate code for a big endian target. This is the default for HP-UX@.
12830 @item -mlittle-endian
12831 @opindex mlittle-endian
12832 Generate code for a little endian target. This is the default for AIX5
12838 @opindex mno-gnu-as
12839 Generate (or don't) code for the GNU assembler. This is the default.
12840 @c Also, this is the default if the configure option @option{--with-gnu-as}
12846 @opindex mno-gnu-ld
12847 Generate (or don't) code for the GNU linker. This is the default.
12848 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12853 Generate code that does not use a global pointer register. The result
12854 is not position independent code, and violates the IA-64 ABI@.
12856 @item -mvolatile-asm-stop
12857 @itemx -mno-volatile-asm-stop
12858 @opindex mvolatile-asm-stop
12859 @opindex mno-volatile-asm-stop
12860 Generate (or don't) a stop bit immediately before and after volatile asm
12863 @item -mregister-names
12864 @itemx -mno-register-names
12865 @opindex mregister-names
12866 @opindex mno-register-names
12867 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12868 the stacked registers. This may make assembler output more readable.
12874 Disable (or enable) optimizations that use the small data section. This may
12875 be useful for working around optimizer bugs.
12877 @item -mconstant-gp
12878 @opindex mconstant-gp
12879 Generate code that uses a single constant global pointer value. This is
12880 useful when compiling kernel code.
12884 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12885 This is useful when compiling firmware code.
12887 @item -minline-float-divide-min-latency
12888 @opindex minline-float-divide-min-latency
12889 Generate code for inline divides of floating point values
12890 using the minimum latency algorithm.
12892 @item -minline-float-divide-max-throughput
12893 @opindex minline-float-divide-max-throughput
12894 Generate code for inline divides of floating point values
12895 using the maximum throughput algorithm.
12897 @item -mno-inline-float-divide
12898 @opindex mno-inline-float-divide
12899 Do not generate inline code for divides of floating point values.
12901 @item -minline-int-divide-min-latency
12902 @opindex minline-int-divide-min-latency
12903 Generate code for inline divides of integer values
12904 using the minimum latency algorithm.
12906 @item -minline-int-divide-max-throughput
12907 @opindex minline-int-divide-max-throughput
12908 Generate code for inline divides of integer values
12909 using the maximum throughput algorithm.
12911 @item -mno-inline-int-divide
12912 @opindex mno-inline-int-divide
12913 Do not generate inline code for divides of integer values.
12915 @item -minline-sqrt-min-latency
12916 @opindex minline-sqrt-min-latency
12917 Generate code for inline square roots
12918 using the minimum latency algorithm.
12920 @item -minline-sqrt-max-throughput
12921 @opindex minline-sqrt-max-throughput
12922 Generate code for inline square roots
12923 using the maximum throughput algorithm.
12925 @item -mno-inline-sqrt
12926 @opindex mno-inline-sqrt
12927 Do not generate inline code for sqrt.
12930 @itemx -mno-fused-madd
12931 @opindex mfused-madd
12932 @opindex mno-fused-madd
12933 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12934 instructions. The default is to use these instructions.
12936 @item -mno-dwarf2-asm
12937 @itemx -mdwarf2-asm
12938 @opindex mno-dwarf2-asm
12939 @opindex mdwarf2-asm
12940 Don't (or do) generate assembler code for the DWARF2 line number debugging
12941 info. This may be useful when not using the GNU assembler.
12943 @item -mearly-stop-bits
12944 @itemx -mno-early-stop-bits
12945 @opindex mearly-stop-bits
12946 @opindex mno-early-stop-bits
12947 Allow stop bits to be placed earlier than immediately preceding the
12948 instruction that triggered the stop bit. This can improve instruction
12949 scheduling, but does not always do so.
12951 @item -mfixed-range=@var{register-range}
12952 @opindex mfixed-range
12953 Generate code treating the given register range as fixed registers.
12954 A fixed register is one that the register allocator can not use. This is
12955 useful when compiling kernel code. A register range is specified as
12956 two registers separated by a dash. Multiple register ranges can be
12957 specified separated by a comma.
12959 @item -mtls-size=@var{tls-size}
12961 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12964 @item -mtune=@var{cpu-type}
12966 Tune the instruction scheduling for a particular CPU, Valid values are
12967 itanium, itanium1, merced, itanium2, and mckinley.
12973 Generate code for a 32-bit or 64-bit environment.
12974 The 32-bit environment sets int, long and pointer to 32 bits.
12975 The 64-bit environment sets int to 32 bits and long and pointer
12976 to 64 bits. These are HP-UX specific flags.
12978 @item -mno-sched-br-data-spec
12979 @itemx -msched-br-data-spec
12980 @opindex mno-sched-br-data-spec
12981 @opindex msched-br-data-spec
12982 (Dis/En)able data speculative scheduling before reload.
12983 This will result in generation of the ld.a instructions and
12984 the corresponding check instructions (ld.c / chk.a).
12985 The default is 'disable'.
12987 @item -msched-ar-data-spec
12988 @itemx -mno-sched-ar-data-spec
12989 @opindex msched-ar-data-spec
12990 @opindex mno-sched-ar-data-spec
12991 (En/Dis)able data speculative scheduling after reload.
12992 This will result in generation of the ld.a instructions and
12993 the corresponding check instructions (ld.c / chk.a).
12994 The default is 'enable'.
12996 @item -mno-sched-control-spec
12997 @itemx -msched-control-spec
12998 @opindex mno-sched-control-spec
12999 @opindex msched-control-spec
13000 (Dis/En)able control speculative scheduling. This feature is
13001 available only during region scheduling (i.e.@: before reload).
13002 This will result in generation of the ld.s instructions and
13003 the corresponding check instructions chk.s .
13004 The default is 'disable'.
13006 @item -msched-br-in-data-spec
13007 @itemx -mno-sched-br-in-data-spec
13008 @opindex msched-br-in-data-spec
13009 @opindex mno-sched-br-in-data-spec
13010 (En/Dis)able speculative scheduling of the instructions that
13011 are dependent on the data speculative loads before reload.
13012 This is effective only with @option{-msched-br-data-spec} enabled.
13013 The default is 'enable'.
13015 @item -msched-ar-in-data-spec
13016 @itemx -mno-sched-ar-in-data-spec
13017 @opindex msched-ar-in-data-spec
13018 @opindex mno-sched-ar-in-data-spec
13019 (En/Dis)able speculative scheduling of the instructions that
13020 are dependent on the data speculative loads after reload.
13021 This is effective only with @option{-msched-ar-data-spec} enabled.
13022 The default is 'enable'.
13024 @item -msched-in-control-spec
13025 @itemx -mno-sched-in-control-spec
13026 @opindex msched-in-control-spec
13027 @opindex mno-sched-in-control-spec
13028 (En/Dis)able speculative scheduling of the instructions that
13029 are dependent on the control speculative loads.
13030 This is effective only with @option{-msched-control-spec} enabled.
13031 The default is 'enable'.
13033 @item -mno-sched-prefer-non-data-spec-insns
13034 @itemx -msched-prefer-non-data-spec-insns
13035 @opindex mno-sched-prefer-non-data-spec-insns
13036 @opindex msched-prefer-non-data-spec-insns
13037 If enabled, data speculative instructions will be chosen for schedule
13038 only if there are no other choices at the moment. This will make
13039 the use of the data speculation much more conservative.
13040 The default is 'disable'.
13042 @item -mno-sched-prefer-non-control-spec-insns
13043 @itemx -msched-prefer-non-control-spec-insns
13044 @opindex mno-sched-prefer-non-control-spec-insns
13045 @opindex msched-prefer-non-control-spec-insns
13046 If enabled, control speculative instructions will be chosen for schedule
13047 only if there are no other choices at the moment. This will make
13048 the use of the control speculation much more conservative.
13049 The default is 'disable'.
13051 @item -mno-sched-count-spec-in-critical-path
13052 @itemx -msched-count-spec-in-critical-path
13053 @opindex mno-sched-count-spec-in-critical-path
13054 @opindex msched-count-spec-in-critical-path
13055 If enabled, speculative dependencies will be considered during
13056 computation of the instructions priorities. This will make the use of the
13057 speculation a bit more conservative.
13058 The default is 'disable'.
13060 @item -msched-spec-ldc
13061 @opindex msched-spec-ldc
13062 Use a simple data speculation check. This option is on by default.
13064 @item -msched-control-spec-ldc
13065 @opindex msched-spec-ldc
13066 Use a simple check for control speculation. This option is on by default.
13068 @item -msched-stop-bits-after-every-cycle
13069 @opindex msched-stop-bits-after-every-cycle
13070 Place a stop bit after every cycle when scheduling. This option is on
13073 @item -msched-fp-mem-deps-zero-cost
13074 @opindex msched-fp-mem-deps-zero-cost
13075 Assume that floating-point stores and loads are not likely to cause a conflict
13076 when placed into the same instruction group. This option is disabled by
13079 @item -msel-sched-dont-check-control-spec
13080 @opindex msel-sched-dont-check-control-spec
13081 Generate checks for control speculation in selective scheduling.
13082 This flag is disabled by default.
13084 @item -msched-max-memory-insns=@var{max-insns}
13085 @opindex msched-max-memory-insns
13086 Limit on the number of memory insns per instruction group, giving lower
13087 priority to subsequent memory insns attempting to schedule in the same
13088 instruction group. Frequently useful to prevent cache bank conflicts.
13089 The default value is 1.
13091 @item -msched-max-memory-insns-hard-limit
13092 @opindex msched-max-memory-insns-hard-limit
13093 Disallow more than `msched-max-memory-insns' in instruction group.
13094 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
13095 when limit is reached but may still schedule memory operations.
13099 @node IA-64/VMS Options
13100 @subsection IA-64/VMS Options
13102 These @samp{-m} options are defined for the IA-64/VMS implementations:
13105 @item -mvms-return-codes
13106 @opindex mvms-return-codes
13107 Return VMS condition codes from main. The default is to return POSIX
13108 style condition (e.g.@ error) codes.
13110 @item -mdebug-main=@var{prefix}
13111 @opindex mdebug-main=@var{prefix}
13112 Flag the first routine whose name starts with @var{prefix} as the main
13113 routine for the debugger.
13117 Default to 64bit memory allocation routines.
13121 @subsection LM32 Options
13122 @cindex LM32 options
13124 These @option{-m} options are defined for the Lattice Mico32 architecture:
13127 @item -mbarrel-shift-enabled
13128 @opindex mbarrel-shift-enabled
13129 Enable barrel-shift instructions.
13131 @item -mdivide-enabled
13132 @opindex mdivide-enabled
13133 Enable divide and modulus instructions.
13135 @item -mmultiply-enabled
13136 @opindex multiply-enabled
13137 Enable multiply instructions.
13139 @item -msign-extend-enabled
13140 @opindex msign-extend-enabled
13141 Enable sign extend instructions.
13143 @item -muser-enabled
13144 @opindex muser-enabled
13145 Enable user-defined instructions.
13150 @subsection M32C Options
13151 @cindex M32C options
13154 @item -mcpu=@var{name}
13156 Select the CPU for which code is generated. @var{name} may be one of
13157 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
13158 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
13159 the M32C/80 series.
13163 Specifies that the program will be run on the simulator. This causes
13164 an alternate runtime library to be linked in which supports, for
13165 example, file I/O@. You must not use this option when generating
13166 programs that will run on real hardware; you must provide your own
13167 runtime library for whatever I/O functions are needed.
13169 @item -memregs=@var{number}
13171 Specifies the number of memory-based pseudo-registers GCC will use
13172 during code generation. These pseudo-registers will be used like real
13173 registers, so there is a tradeoff between GCC's ability to fit the
13174 code into available registers, and the performance penalty of using
13175 memory instead of registers. Note that all modules in a program must
13176 be compiled with the same value for this option. Because of that, you
13177 must not use this option with the default runtime libraries gcc
13182 @node M32R/D Options
13183 @subsection M32R/D Options
13184 @cindex M32R/D options
13186 These @option{-m} options are defined for Renesas M32R/D architectures:
13191 Generate code for the M32R/2@.
13195 Generate code for the M32R/X@.
13199 Generate code for the M32R@. This is the default.
13201 @item -mmodel=small
13202 @opindex mmodel=small
13203 Assume all objects live in the lower 16MB of memory (so that their addresses
13204 can be loaded with the @code{ld24} instruction), and assume all subroutines
13205 are reachable with the @code{bl} instruction.
13206 This is the default.
13208 The addressability of a particular object can be set with the
13209 @code{model} attribute.
13211 @item -mmodel=medium
13212 @opindex mmodel=medium
13213 Assume objects may be anywhere in the 32-bit address space (the compiler
13214 will generate @code{seth/add3} instructions to load their addresses), and
13215 assume all subroutines are reachable with the @code{bl} instruction.
13217 @item -mmodel=large
13218 @opindex mmodel=large
13219 Assume objects may be anywhere in the 32-bit address space (the compiler
13220 will generate @code{seth/add3} instructions to load their addresses), and
13221 assume subroutines may not be reachable with the @code{bl} instruction
13222 (the compiler will generate the much slower @code{seth/add3/jl}
13223 instruction sequence).
13226 @opindex msdata=none
13227 Disable use of the small data area. Variables will be put into
13228 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
13229 @code{section} attribute has been specified).
13230 This is the default.
13232 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
13233 Objects may be explicitly put in the small data area with the
13234 @code{section} attribute using one of these sections.
13236 @item -msdata=sdata
13237 @opindex msdata=sdata
13238 Put small global and static data in the small data area, but do not
13239 generate special code to reference them.
13242 @opindex msdata=use
13243 Put small global and static data in the small data area, and generate
13244 special instructions to reference them.
13248 @cindex smaller data references
13249 Put global and static objects less than or equal to @var{num} bytes
13250 into the small data or bss sections instead of the normal data or bss
13251 sections. The default value of @var{num} is 8.
13252 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
13253 for this option to have any effect.
13255 All modules should be compiled with the same @option{-G @var{num}} value.
13256 Compiling with different values of @var{num} may or may not work; if it
13257 doesn't the linker will give an error message---incorrect code will not be
13262 Makes the M32R specific code in the compiler display some statistics
13263 that might help in debugging programs.
13265 @item -malign-loops
13266 @opindex malign-loops
13267 Align all loops to a 32-byte boundary.
13269 @item -mno-align-loops
13270 @opindex mno-align-loops
13271 Do not enforce a 32-byte alignment for loops. This is the default.
13273 @item -missue-rate=@var{number}
13274 @opindex missue-rate=@var{number}
13275 Issue @var{number} instructions per cycle. @var{number} can only be 1
13278 @item -mbranch-cost=@var{number}
13279 @opindex mbranch-cost=@var{number}
13280 @var{number} can only be 1 or 2. If it is 1 then branches will be
13281 preferred over conditional code, if it is 2, then the opposite will
13284 @item -mflush-trap=@var{number}
13285 @opindex mflush-trap=@var{number}
13286 Specifies the trap number to use to flush the cache. The default is
13287 12. Valid numbers are between 0 and 15 inclusive.
13289 @item -mno-flush-trap
13290 @opindex mno-flush-trap
13291 Specifies that the cache cannot be flushed by using a trap.
13293 @item -mflush-func=@var{name}
13294 @opindex mflush-func=@var{name}
13295 Specifies the name of the operating system function to call to flush
13296 the cache. The default is @emph{_flush_cache}, but a function call
13297 will only be used if a trap is not available.
13299 @item -mno-flush-func
13300 @opindex mno-flush-func
13301 Indicates that there is no OS function for flushing the cache.
13305 @node M680x0 Options
13306 @subsection M680x0 Options
13307 @cindex M680x0 options
13309 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
13310 The default settings depend on which architecture was selected when
13311 the compiler was configured; the defaults for the most common choices
13315 @item -march=@var{arch}
13317 Generate code for a specific M680x0 or ColdFire instruction set
13318 architecture. Permissible values of @var{arch} for M680x0
13319 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
13320 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
13321 architectures are selected according to Freescale's ISA classification
13322 and the permissible values are: @samp{isaa}, @samp{isaaplus},
13323 @samp{isab} and @samp{isac}.
13325 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
13326 code for a ColdFire target. The @var{arch} in this macro is one of the
13327 @option{-march} arguments given above.
13329 When used together, @option{-march} and @option{-mtune} select code
13330 that runs on a family of similar processors but that is optimized
13331 for a particular microarchitecture.
13333 @item -mcpu=@var{cpu}
13335 Generate code for a specific M680x0 or ColdFire processor.
13336 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
13337 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
13338 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
13339 below, which also classifies the CPUs into families:
13341 @multitable @columnfractions 0.20 0.80
13342 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
13343 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
13344 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
13345 @item @samp{5206e} @tab @samp{5206e}
13346 @item @samp{5208} @tab @samp{5207} @samp{5208}
13347 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
13348 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
13349 @item @samp{5216} @tab @samp{5214} @samp{5216}
13350 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
13351 @item @samp{5225} @tab @samp{5224} @samp{5225}
13352 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
13353 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
13354 @item @samp{5249} @tab @samp{5249}
13355 @item @samp{5250} @tab @samp{5250}
13356 @item @samp{5271} @tab @samp{5270} @samp{5271}
13357 @item @samp{5272} @tab @samp{5272}
13358 @item @samp{5275} @tab @samp{5274} @samp{5275}
13359 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
13360 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
13361 @item @samp{5307} @tab @samp{5307}
13362 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
13363 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
13364 @item @samp{5407} @tab @samp{5407}
13365 @item @samp{5475} @tab @samp{5470} @samp{5471} @samp{5472} @samp{5473} @samp{5474} @samp{5475} @samp{547x} @samp{5480} @samp{5481} @samp{5482} @samp{5483} @samp{5484} @samp{5485}
13368 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
13369 @var{arch} is compatible with @var{cpu}. Other combinations of
13370 @option{-mcpu} and @option{-march} are rejected.
13372 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
13373 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
13374 where the value of @var{family} is given by the table above.
13376 @item -mtune=@var{tune}
13378 Tune the code for a particular microarchitecture, within the
13379 constraints set by @option{-march} and @option{-mcpu}.
13380 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
13381 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
13382 and @samp{cpu32}. The ColdFire microarchitectures
13383 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
13385 You can also use @option{-mtune=68020-40} for code that needs
13386 to run relatively well on 68020, 68030 and 68040 targets.
13387 @option{-mtune=68020-60} is similar but includes 68060 targets
13388 as well. These two options select the same tuning decisions as
13389 @option{-m68020-40} and @option{-m68020-60} respectively.
13391 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
13392 when tuning for 680x0 architecture @var{arch}. It also defines
13393 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
13394 option is used. If gcc is tuning for a range of architectures,
13395 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
13396 it defines the macros for every architecture in the range.
13398 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
13399 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
13400 of the arguments given above.
13406 Generate output for a 68000. This is the default
13407 when the compiler is configured for 68000-based systems.
13408 It is equivalent to @option{-march=68000}.
13410 Use this option for microcontrollers with a 68000 or EC000 core,
13411 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
13415 Generate output for a 68010. This is the default
13416 when the compiler is configured for 68010-based systems.
13417 It is equivalent to @option{-march=68010}.
13423 Generate output for a 68020. This is the default
13424 when the compiler is configured for 68020-based systems.
13425 It is equivalent to @option{-march=68020}.
13429 Generate output for a 68030. This is the default when the compiler is
13430 configured for 68030-based systems. It is equivalent to
13431 @option{-march=68030}.
13435 Generate output for a 68040. This is the default when the compiler is
13436 configured for 68040-based systems. It is equivalent to
13437 @option{-march=68040}.
13439 This option inhibits the use of 68881/68882 instructions that have to be
13440 emulated by software on the 68040. Use this option if your 68040 does not
13441 have code to emulate those instructions.
13445 Generate output for a 68060. This is the default when the compiler is
13446 configured for 68060-based systems. It is equivalent to
13447 @option{-march=68060}.
13449 This option inhibits the use of 68020 and 68881/68882 instructions that
13450 have to be emulated by software on the 68060. Use this option if your 68060
13451 does not have code to emulate those instructions.
13455 Generate output for a CPU32. This is the default
13456 when the compiler is configured for CPU32-based systems.
13457 It is equivalent to @option{-march=cpu32}.
13459 Use this option for microcontrollers with a
13460 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
13461 68336, 68340, 68341, 68349 and 68360.
13465 Generate output for a 520X ColdFire CPU@. This is the default
13466 when the compiler is configured for 520X-based systems.
13467 It is equivalent to @option{-mcpu=5206}, and is now deprecated
13468 in favor of that option.
13470 Use this option for microcontroller with a 5200 core, including
13471 the MCF5202, MCF5203, MCF5204 and MCF5206.
13475 Generate output for a 5206e ColdFire CPU@. The option is now
13476 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13480 Generate output for a member of the ColdFire 528X family.
13481 The option is now deprecated in favor of the equivalent
13482 @option{-mcpu=528x}.
13486 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
13487 in favor of the equivalent @option{-mcpu=5307}.
13491 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
13492 in favor of the equivalent @option{-mcpu=5407}.
13496 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
13497 This includes use of hardware floating point instructions.
13498 The option is equivalent to @option{-mcpu=547x}, and is now
13499 deprecated in favor of that option.
13503 Generate output for a 68040, without using any of the new instructions.
13504 This results in code which can run relatively efficiently on either a
13505 68020/68881 or a 68030 or a 68040. The generated code does use the
13506 68881 instructions that are emulated on the 68040.
13508 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
13512 Generate output for a 68060, without using any of the new instructions.
13513 This results in code which can run relatively efficiently on either a
13514 68020/68881 or a 68030 or a 68040. The generated code does use the
13515 68881 instructions that are emulated on the 68060.
13517 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
13521 @opindex mhard-float
13523 Generate floating-point instructions. This is the default for 68020
13524 and above, and for ColdFire devices that have an FPU@. It defines the
13525 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
13526 on ColdFire targets.
13529 @opindex msoft-float
13530 Do not generate floating-point instructions; use library calls instead.
13531 This is the default for 68000, 68010, and 68832 targets. It is also
13532 the default for ColdFire devices that have no FPU.
13538 Generate (do not generate) ColdFire hardware divide and remainder
13539 instructions. If @option{-march} is used without @option{-mcpu},
13540 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13541 architectures. Otherwise, the default is taken from the target CPU
13542 (either the default CPU, or the one specified by @option{-mcpu}). For
13543 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13544 @option{-mcpu=5206e}.
13546 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13550 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13551 Additionally, parameters passed on the stack are also aligned to a
13552 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13556 Do not consider type @code{int} to be 16 bits wide. This is the default.
13559 @itemx -mno-bitfield
13560 @opindex mnobitfield
13561 @opindex mno-bitfield
13562 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13563 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13567 Do use the bit-field instructions. The @option{-m68020} option implies
13568 @option{-mbitfield}. This is the default if you use a configuration
13569 designed for a 68020.
13573 Use a different function-calling convention, in which functions
13574 that take a fixed number of arguments return with the @code{rtd}
13575 instruction, which pops their arguments while returning. This
13576 saves one instruction in the caller since there is no need to pop
13577 the arguments there.
13579 This calling convention is incompatible with the one normally
13580 used on Unix, so you cannot use it if you need to call libraries
13581 compiled with the Unix compiler.
13583 Also, you must provide function prototypes for all functions that
13584 take variable numbers of arguments (including @code{printf});
13585 otherwise incorrect code will be generated for calls to those
13588 In addition, seriously incorrect code will result if you call a
13589 function with too many arguments. (Normally, extra arguments are
13590 harmlessly ignored.)
13592 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13593 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13597 Do not use the calling conventions selected by @option{-mrtd}.
13598 This is the default.
13601 @itemx -mno-align-int
13602 @opindex malign-int
13603 @opindex mno-align-int
13604 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13605 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13606 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13607 Aligning variables on 32-bit boundaries produces code that runs somewhat
13608 faster on processors with 32-bit busses at the expense of more memory.
13610 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13611 align structures containing the above types differently than
13612 most published application binary interface specifications for the m68k.
13616 Use the pc-relative addressing mode of the 68000 directly, instead of
13617 using a global offset table. At present, this option implies @option{-fpic},
13618 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13619 not presently supported with @option{-mpcrel}, though this could be supported for
13620 68020 and higher processors.
13622 @item -mno-strict-align
13623 @itemx -mstrict-align
13624 @opindex mno-strict-align
13625 @opindex mstrict-align
13626 Do not (do) assume that unaligned memory references will be handled by
13630 Generate code that allows the data segment to be located in a different
13631 area of memory from the text segment. This allows for execute in place in
13632 an environment without virtual memory management. This option implies
13635 @item -mno-sep-data
13636 Generate code that assumes that the data segment follows the text segment.
13637 This is the default.
13639 @item -mid-shared-library
13640 Generate code that supports shared libraries via the library ID method.
13641 This allows for execute in place and shared libraries in an environment
13642 without virtual memory management. This option implies @option{-fPIC}.
13644 @item -mno-id-shared-library
13645 Generate code that doesn't assume ID based shared libraries are being used.
13646 This is the default.
13648 @item -mshared-library-id=n
13649 Specified the identification number of the ID based shared library being
13650 compiled. Specifying a value of 0 will generate more compact code, specifying
13651 other values will force the allocation of that number to the current
13652 library but is no more space or time efficient than omitting this option.
13658 When generating position-independent code for ColdFire, generate code
13659 that works if the GOT has more than 8192 entries. This code is
13660 larger and slower than code generated without this option. On M680x0
13661 processors, this option is not needed; @option{-fPIC} suffices.
13663 GCC normally uses a single instruction to load values from the GOT@.
13664 While this is relatively efficient, it only works if the GOT
13665 is smaller than about 64k. Anything larger causes the linker
13666 to report an error such as:
13668 @cindex relocation truncated to fit (ColdFire)
13670 relocation truncated to fit: R_68K_GOT16O foobar
13673 If this happens, you should recompile your code with @option{-mxgot}.
13674 It should then work with very large GOTs. However, code generated with
13675 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13676 the value of a global symbol.
13678 Note that some linkers, including newer versions of the GNU linker,
13679 can create multiple GOTs and sort GOT entries. If you have such a linker,
13680 you should only need to use @option{-mxgot} when compiling a single
13681 object file that accesses more than 8192 GOT entries. Very few do.
13683 These options have no effect unless GCC is generating
13684 position-independent code.
13688 @node M68hc1x Options
13689 @subsection M68hc1x Options
13690 @cindex M68hc1x options
13692 These are the @samp{-m} options defined for the 68hc11 and 68hc12
13693 microcontrollers. The default values for these options depends on
13694 which style of microcontroller was selected when the compiler was configured;
13695 the defaults for the most common choices are given below.
13702 Generate output for a 68HC11. This is the default
13703 when the compiler is configured for 68HC11-based systems.
13709 Generate output for a 68HC12. This is the default
13710 when the compiler is configured for 68HC12-based systems.
13716 Generate output for a 68HCS12.
13718 @item -mauto-incdec
13719 @opindex mauto-incdec
13720 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
13727 Enable the use of 68HC12 min and max instructions.
13730 @itemx -mno-long-calls
13731 @opindex mlong-calls
13732 @opindex mno-long-calls
13733 Treat all calls as being far away (near). If calls are assumed to be
13734 far away, the compiler will use the @code{call} instruction to
13735 call a function and the @code{rtc} instruction for returning.
13739 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13741 @item -msoft-reg-count=@var{count}
13742 @opindex msoft-reg-count
13743 Specify the number of pseudo-soft registers which are used for the
13744 code generation. The maximum number is 32. Using more pseudo-soft
13745 register may or may not result in better code depending on the program.
13746 The default is 4 for 68HC11 and 2 for 68HC12.
13750 @node MCore Options
13751 @subsection MCore Options
13752 @cindex MCore options
13754 These are the @samp{-m} options defined for the Motorola M*Core
13760 @itemx -mno-hardlit
13762 @opindex mno-hardlit
13763 Inline constants into the code stream if it can be done in two
13764 instructions or less.
13770 Use the divide instruction. (Enabled by default).
13772 @item -mrelax-immediate
13773 @itemx -mno-relax-immediate
13774 @opindex mrelax-immediate
13775 @opindex mno-relax-immediate
13776 Allow arbitrary sized immediates in bit operations.
13778 @item -mwide-bitfields
13779 @itemx -mno-wide-bitfields
13780 @opindex mwide-bitfields
13781 @opindex mno-wide-bitfields
13782 Always treat bit-fields as int-sized.
13784 @item -m4byte-functions
13785 @itemx -mno-4byte-functions
13786 @opindex m4byte-functions
13787 @opindex mno-4byte-functions
13788 Force all functions to be aligned to a four byte boundary.
13790 @item -mcallgraph-data
13791 @itemx -mno-callgraph-data
13792 @opindex mcallgraph-data
13793 @opindex mno-callgraph-data
13794 Emit callgraph information.
13797 @itemx -mno-slow-bytes
13798 @opindex mslow-bytes
13799 @opindex mno-slow-bytes
13800 Prefer word access when reading byte quantities.
13802 @item -mlittle-endian
13803 @itemx -mbig-endian
13804 @opindex mlittle-endian
13805 @opindex mbig-endian
13806 Generate code for a little endian target.
13812 Generate code for the 210 processor.
13816 Assume that run-time support has been provided and so omit the
13817 simulator library (@file{libsim.a)} from the linker command line.
13819 @item -mstack-increment=@var{size}
13820 @opindex mstack-increment
13821 Set the maximum amount for a single stack increment operation. Large
13822 values can increase the speed of programs which contain functions
13823 that need a large amount of stack space, but they can also trigger a
13824 segmentation fault if the stack is extended too much. The default
13830 @subsection MeP Options
13831 @cindex MeP options
13837 Enables the @code{abs} instruction, which is the absolute difference
13838 between two registers.
13842 Enables all the optional instructions - average, multiply, divide, bit
13843 operations, leading zero, absolute difference, min/max, clip, and
13849 Enables the @code{ave} instruction, which computes the average of two
13852 @item -mbased=@var{n}
13854 Variables of size @var{n} bytes or smaller will be placed in the
13855 @code{.based} section by default. Based variables use the @code{$tp}
13856 register as a base register, and there is a 128 byte limit to the
13857 @code{.based} section.
13861 Enables the bit operation instructions - bit test (@code{btstm}), set
13862 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13863 test-and-set (@code{tas}).
13865 @item -mc=@var{name}
13867 Selects which section constant data will be placed in. @var{name} may
13868 be @code{tiny}, @code{near}, or @code{far}.
13872 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13873 useful unless you also provide @code{-mminmax}.
13875 @item -mconfig=@var{name}
13877 Selects one of the build-in core configurations. Each MeP chip has
13878 one or more modules in it; each module has a core CPU and a variety of
13879 coprocessors, optional instructions, and peripherals. The
13880 @code{MeP-Integrator} tool, not part of GCC, provides these
13881 configurations through this option; using this option is the same as
13882 using all the corresponding command line options. The default
13883 configuration is @code{default}.
13887 Enables the coprocessor instructions. By default, this is a 32-bit
13888 coprocessor. Note that the coprocessor is normally enabled via the
13889 @code{-mconfig=} option.
13893 Enables the 32-bit coprocessor's instructions.
13897 Enables the 64-bit coprocessor's instructions.
13901 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13905 Causes constant variables to be placed in the @code{.near} section.
13909 Enables the @code{div} and @code{divu} instructions.
13913 Generate big-endian code.
13917 Generate little-endian code.
13919 @item -mio-volatile
13920 @opindex mio-volatile
13921 Tells the compiler that any variable marked with the @code{io}
13922 attribute is to be considered volatile.
13926 Causes variables to be assigned to the @code{.far} section by default.
13930 Enables the @code{leadz} (leading zero) instruction.
13934 Causes variables to be assigned to the @code{.near} section by default.
13938 Enables the @code{min} and @code{max} instructions.
13942 Enables the multiplication and multiply-accumulate instructions.
13946 Disables all the optional instructions enabled by @code{-mall-opts}.
13950 Enables the @code{repeat} and @code{erepeat} instructions, used for
13951 low-overhead looping.
13955 Causes all variables to default to the @code{.tiny} section. Note
13956 that there is a 65536 byte limit to this section. Accesses to these
13957 variables use the @code{%gp} base register.
13961 Enables the saturation instructions. Note that the compiler does not
13962 currently generate these itself, but this option is included for
13963 compatibility with other tools, like @code{as}.
13967 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13971 Link the simulator runtime libraries.
13975 Link the simulator runtime libraries, excluding built-in support
13976 for reset and exception vectors and tables.
13980 Causes all functions to default to the @code{.far} section. Without
13981 this option, functions default to the @code{.near} section.
13983 @item -mtiny=@var{n}
13985 Variables that are @var{n} bytes or smaller will be allocated to the
13986 @code{.tiny} section. These variables use the @code{$gp} base
13987 register. The default for this option is 4, but note that there's a
13988 65536 byte limit to the @code{.tiny} section.
13992 @node MicroBlaze Options
13993 @subsection MicroBlaze Options
13994 @cindex MicroBlaze Options
13999 @opindex msoft-float
14000 Use software emulation for floating point (default).
14003 @opindex mhard-float
14004 Use hardware floating point instructions.
14008 Do not optimize block moves, use @code{memcpy}.
14010 @item -mno-clearbss
14011 @opindex mno-clearbss
14012 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
14014 @item -mcpu=@var{cpu-type}
14016 Use features of and schedule code for given CPU.
14017 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
14018 where @var{X} is a major version, @var{YY} is the minor version, and
14019 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
14020 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
14022 @item -mxl-soft-mul
14023 @opindex mxl-soft-mul
14024 Use software multiply emulation (default).
14026 @item -mxl-soft-div
14027 @opindex mxl-soft-div
14028 Use software emulation for divides (default).
14030 @item -mxl-barrel-shift
14031 @opindex mxl-barrel-shift
14032 Use the hardware barrel shifter.
14034 @item -mxl-pattern-compare
14035 @opindex mxl-pattern-compare
14036 Use pattern compare instructions.
14038 @item -msmall-divides
14039 @opindex msmall-divides
14040 Use table lookup optimization for small signed integer divisions.
14042 @item -mxl-stack-check
14043 @opindex mxl-stack-check
14044 This option is deprecated. Use -fstack-check instead.
14047 @opindex mxl-gp-opt
14048 Use GP relative sdata/sbss sections.
14050 @item -mxl-multiply-high
14051 @opindex mxl-multiply-high
14052 Use multiply high instructions for high part of 32x32 multiply.
14054 @item -mxl-float-convert
14055 @opindex mxl-float-convert
14056 Use hardware floating point conversion instructions.
14058 @item -mxl-float-sqrt
14059 @opindex mxl-float-sqrt
14060 Use hardware floating point square root instruction.
14062 @item -mxl-mode-@var{app-model}
14063 Select application model @var{app-model}. Valid models are
14066 normal executable (default), uses startup code @file{crt0.o}.
14069 for use with Xilinx Microprocessor Debugger (XMD) based
14070 software intrusive debug agent called xmdstub. This uses startup file
14071 @file{crt1.o} and sets the start address of the program to be 0x800.
14074 for applications that are loaded using a bootloader.
14075 This model uses startup file @file{crt2.o} which does not contain a processor
14076 reset vector handler. This is suitable for transferring control on a
14077 processor reset to the bootloader rather than the application.
14080 for applications that do not require any of the
14081 MicroBlaze vectors. This option may be useful for applications running
14082 within a monitoring application. This model uses @file{crt3.o} as a startup file.
14085 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
14086 @option{-mxl-mode-@var{app-model}}.
14091 @subsection MIPS Options
14092 @cindex MIPS options
14098 Generate big-endian code.
14102 Generate little-endian code. This is the default for @samp{mips*el-*-*}
14105 @item -march=@var{arch}
14107 Generate code that will run on @var{arch}, which can be the name of a
14108 generic MIPS ISA, or the name of a particular processor.
14110 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
14111 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
14112 The processor names are:
14113 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
14114 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
14115 @samp{5kc}, @samp{5kf},
14117 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
14118 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
14119 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
14120 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
14121 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
14122 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
14126 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
14127 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
14128 @samp{rm7000}, @samp{rm9000},
14129 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
14132 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
14133 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
14135 The special value @samp{from-abi} selects the
14136 most compatible architecture for the selected ABI (that is,
14137 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
14139 Native Linux/GNU toolchains also support the value @samp{native},
14140 which selects the best architecture option for the host processor.
14141 @option{-march=native} has no effect if GCC does not recognize
14144 In processor names, a final @samp{000} can be abbreviated as @samp{k}
14145 (for example, @samp{-march=r2k}). Prefixes are optional, and
14146 @samp{vr} may be written @samp{r}.
14148 Names of the form @samp{@var{n}f2_1} refer to processors with
14149 FPUs clocked at half the rate of the core, names of the form
14150 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
14151 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
14152 processors with FPUs clocked a ratio of 3:2 with respect to the core.
14153 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
14154 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
14155 accepted as synonyms for @samp{@var{n}f1_1}.
14157 GCC defines two macros based on the value of this option. The first
14158 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
14159 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
14160 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
14161 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
14162 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
14164 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
14165 above. In other words, it will have the full prefix and will not
14166 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
14167 the macro names the resolved architecture (either @samp{"mips1"} or
14168 @samp{"mips3"}). It names the default architecture when no
14169 @option{-march} option is given.
14171 @item -mtune=@var{arch}
14173 Optimize for @var{arch}. Among other things, this option controls
14174 the way instructions are scheduled, and the perceived cost of arithmetic
14175 operations. The list of @var{arch} values is the same as for
14178 When this option is not used, GCC will optimize for the processor
14179 specified by @option{-march}. By using @option{-march} and
14180 @option{-mtune} together, it is possible to generate code that will
14181 run on a family of processors, but optimize the code for one
14182 particular member of that family.
14184 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
14185 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
14186 @samp{-march} ones described above.
14190 Equivalent to @samp{-march=mips1}.
14194 Equivalent to @samp{-march=mips2}.
14198 Equivalent to @samp{-march=mips3}.
14202 Equivalent to @samp{-march=mips4}.
14206 Equivalent to @samp{-march=mips32}.
14210 Equivalent to @samp{-march=mips32r2}.
14214 Equivalent to @samp{-march=mips64}.
14218 Equivalent to @samp{-march=mips64r2}.
14223 @opindex mno-mips16
14224 Generate (do not generate) MIPS16 code. If GCC is targetting a
14225 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
14227 MIPS16 code generation can also be controlled on a per-function basis
14228 by means of @code{mips16} and @code{nomips16} attributes.
14229 @xref{Function Attributes}, for more information.
14231 @item -mflip-mips16
14232 @opindex mflip-mips16
14233 Generate MIPS16 code on alternating functions. This option is provided
14234 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
14235 not intended for ordinary use in compiling user code.
14237 @item -minterlink-mips16
14238 @itemx -mno-interlink-mips16
14239 @opindex minterlink-mips16
14240 @opindex mno-interlink-mips16
14241 Require (do not require) that non-MIPS16 code be link-compatible with
14244 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
14245 it must either use a call or an indirect jump. @option{-minterlink-mips16}
14246 therefore disables direct jumps unless GCC knows that the target of the
14247 jump is not MIPS16.
14259 Generate code for the given ABI@.
14261 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
14262 generates 64-bit code when you select a 64-bit architecture, but you
14263 can use @option{-mgp32} to get 32-bit code instead.
14265 For information about the O64 ABI, see
14266 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
14268 GCC supports a variant of the o32 ABI in which floating-point registers
14269 are 64 rather than 32 bits wide. You can select this combination with
14270 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
14271 and @samp{mfhc1} instructions and is therefore only supported for
14272 MIPS32R2 processors.
14274 The register assignments for arguments and return values remain the
14275 same, but each scalar value is passed in a single 64-bit register
14276 rather than a pair of 32-bit registers. For example, scalar
14277 floating-point values are returned in @samp{$f0} only, not a
14278 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
14279 remains the same, but all 64 bits are saved.
14282 @itemx -mno-abicalls
14284 @opindex mno-abicalls
14285 Generate (do not generate) code that is suitable for SVR4-style
14286 dynamic objects. @option{-mabicalls} is the default for SVR4-based
14291 Generate (do not generate) code that is fully position-independent,
14292 and that can therefore be linked into shared libraries. This option
14293 only affects @option{-mabicalls}.
14295 All @option{-mabicalls} code has traditionally been position-independent,
14296 regardless of options like @option{-fPIC} and @option{-fpic}. However,
14297 as an extension, the GNU toolchain allows executables to use absolute
14298 accesses for locally-binding symbols. It can also use shorter GP
14299 initialization sequences and generate direct calls to locally-defined
14300 functions. This mode is selected by @option{-mno-shared}.
14302 @option{-mno-shared} depends on binutils 2.16 or higher and generates
14303 objects that can only be linked by the GNU linker. However, the option
14304 does not affect the ABI of the final executable; it only affects the ABI
14305 of relocatable objects. Using @option{-mno-shared} will generally make
14306 executables both smaller and quicker.
14308 @option{-mshared} is the default.
14314 Assume (do not assume) that the static and dynamic linkers
14315 support PLTs and copy relocations. This option only affects
14316 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
14317 has no effect without @samp{-msym32}.
14319 You can make @option{-mplt} the default by configuring
14320 GCC with @option{--with-mips-plt}. The default is
14321 @option{-mno-plt} otherwise.
14327 Lift (do not lift) the usual restrictions on the size of the global
14330 GCC normally uses a single instruction to load values from the GOT@.
14331 While this is relatively efficient, it will only work if the GOT
14332 is smaller than about 64k. Anything larger will cause the linker
14333 to report an error such as:
14335 @cindex relocation truncated to fit (MIPS)
14337 relocation truncated to fit: R_MIPS_GOT16 foobar
14340 If this happens, you should recompile your code with @option{-mxgot}.
14341 It should then work with very large GOTs, although it will also be
14342 less efficient, since it will take three instructions to fetch the
14343 value of a global symbol.
14345 Note that some linkers can create multiple GOTs. If you have such a
14346 linker, you should only need to use @option{-mxgot} when a single object
14347 file accesses more than 64k's worth of GOT entries. Very few do.
14349 These options have no effect unless GCC is generating position
14354 Assume that general-purpose registers are 32 bits wide.
14358 Assume that general-purpose registers are 64 bits wide.
14362 Assume that floating-point registers are 32 bits wide.
14366 Assume that floating-point registers are 64 bits wide.
14369 @opindex mhard-float
14370 Use floating-point coprocessor instructions.
14373 @opindex msoft-float
14374 Do not use floating-point coprocessor instructions. Implement
14375 floating-point calculations using library calls instead.
14377 @item -msingle-float
14378 @opindex msingle-float
14379 Assume that the floating-point coprocessor only supports single-precision
14382 @item -mdouble-float
14383 @opindex mdouble-float
14384 Assume that the floating-point coprocessor supports double-precision
14385 operations. This is the default.
14391 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
14392 implement atomic memory built-in functions. When neither option is
14393 specified, GCC will use the instructions if the target architecture
14396 @option{-mllsc} is useful if the runtime environment can emulate the
14397 instructions and @option{-mno-llsc} can be useful when compiling for
14398 nonstandard ISAs. You can make either option the default by
14399 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
14400 respectively. @option{--with-llsc} is the default for some
14401 configurations; see the installation documentation for details.
14407 Use (do not use) revision 1 of the MIPS DSP ASE@.
14408 @xref{MIPS DSP Built-in Functions}. This option defines the
14409 preprocessor macro @samp{__mips_dsp}. It also defines
14410 @samp{__mips_dsp_rev} to 1.
14416 Use (do not use) revision 2 of the MIPS DSP ASE@.
14417 @xref{MIPS DSP Built-in Functions}. This option defines the
14418 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
14419 It also defines @samp{__mips_dsp_rev} to 2.
14422 @itemx -mno-smartmips
14423 @opindex msmartmips
14424 @opindex mno-smartmips
14425 Use (do not use) the MIPS SmartMIPS ASE.
14427 @item -mpaired-single
14428 @itemx -mno-paired-single
14429 @opindex mpaired-single
14430 @opindex mno-paired-single
14431 Use (do not use) paired-single floating-point instructions.
14432 @xref{MIPS Paired-Single Support}. This option requires
14433 hardware floating-point support to be enabled.
14439 Use (do not use) MIPS Digital Media Extension instructions.
14440 This option can only be used when generating 64-bit code and requires
14441 hardware floating-point support to be enabled.
14446 @opindex mno-mips3d
14447 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
14448 The option @option{-mips3d} implies @option{-mpaired-single}.
14454 Use (do not use) MT Multithreading instructions.
14458 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
14459 an explanation of the default and the way that the pointer size is
14464 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
14466 The default size of @code{int}s, @code{long}s and pointers depends on
14467 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
14468 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
14469 32-bit @code{long}s. Pointers are the same size as @code{long}s,
14470 or the same size as integer registers, whichever is smaller.
14476 Assume (do not assume) that all symbols have 32-bit values, regardless
14477 of the selected ABI@. This option is useful in combination with
14478 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
14479 to generate shorter and faster references to symbolic addresses.
14483 Put definitions of externally-visible data in a small data section
14484 if that data is no bigger than @var{num} bytes. GCC can then access
14485 the data more efficiently; see @option{-mgpopt} for details.
14487 The default @option{-G} option depends on the configuration.
14489 @item -mlocal-sdata
14490 @itemx -mno-local-sdata
14491 @opindex mlocal-sdata
14492 @opindex mno-local-sdata
14493 Extend (do not extend) the @option{-G} behavior to local data too,
14494 such as to static variables in C@. @option{-mlocal-sdata} is the
14495 default for all configurations.
14497 If the linker complains that an application is using too much small data,
14498 you might want to try rebuilding the less performance-critical parts with
14499 @option{-mno-local-sdata}. You might also want to build large
14500 libraries with @option{-mno-local-sdata}, so that the libraries leave
14501 more room for the main program.
14503 @item -mextern-sdata
14504 @itemx -mno-extern-sdata
14505 @opindex mextern-sdata
14506 @opindex mno-extern-sdata
14507 Assume (do not assume) that externally-defined data will be in
14508 a small data section if that data is within the @option{-G} limit.
14509 @option{-mextern-sdata} is the default for all configurations.
14511 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
14512 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
14513 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
14514 is placed in a small data section. If @var{Var} is defined by another
14515 module, you must either compile that module with a high-enough
14516 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
14517 definition. If @var{Var} is common, you must link the application
14518 with a high-enough @option{-G} setting.
14520 The easiest way of satisfying these restrictions is to compile
14521 and link every module with the same @option{-G} option. However,
14522 you may wish to build a library that supports several different
14523 small data limits. You can do this by compiling the library with
14524 the highest supported @option{-G} setting and additionally using
14525 @option{-mno-extern-sdata} to stop the library from making assumptions
14526 about externally-defined data.
14532 Use (do not use) GP-relative accesses for symbols that are known to be
14533 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
14534 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
14537 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
14538 might not hold the value of @code{_gp}. For example, if the code is
14539 part of a library that might be used in a boot monitor, programs that
14540 call boot monitor routines will pass an unknown value in @code{$gp}.
14541 (In such situations, the boot monitor itself would usually be compiled
14542 with @option{-G0}.)
14544 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
14545 @option{-mno-extern-sdata}.
14547 @item -membedded-data
14548 @itemx -mno-embedded-data
14549 @opindex membedded-data
14550 @opindex mno-embedded-data
14551 Allocate variables to the read-only data section first if possible, then
14552 next in the small data section if possible, otherwise in data. This gives
14553 slightly slower code than the default, but reduces the amount of RAM required
14554 when executing, and thus may be preferred for some embedded systems.
14556 @item -muninit-const-in-rodata
14557 @itemx -mno-uninit-const-in-rodata
14558 @opindex muninit-const-in-rodata
14559 @opindex mno-uninit-const-in-rodata
14560 Put uninitialized @code{const} variables in the read-only data section.
14561 This option is only meaningful in conjunction with @option{-membedded-data}.
14563 @item -mcode-readable=@var{setting}
14564 @opindex mcode-readable
14565 Specify whether GCC may generate code that reads from executable sections.
14566 There are three possible settings:
14569 @item -mcode-readable=yes
14570 Instructions may freely access executable sections. This is the
14573 @item -mcode-readable=pcrel
14574 MIPS16 PC-relative load instructions can access executable sections,
14575 but other instructions must not do so. This option is useful on 4KSc
14576 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
14577 It is also useful on processors that can be configured to have a dual
14578 instruction/data SRAM interface and that, like the M4K, automatically
14579 redirect PC-relative loads to the instruction RAM.
14581 @item -mcode-readable=no
14582 Instructions must not access executable sections. This option can be
14583 useful on targets that are configured to have a dual instruction/data
14584 SRAM interface but that (unlike the M4K) do not automatically redirect
14585 PC-relative loads to the instruction RAM.
14588 @item -msplit-addresses
14589 @itemx -mno-split-addresses
14590 @opindex msplit-addresses
14591 @opindex mno-split-addresses
14592 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
14593 relocation operators. This option has been superseded by
14594 @option{-mexplicit-relocs} but is retained for backwards compatibility.
14596 @item -mexplicit-relocs
14597 @itemx -mno-explicit-relocs
14598 @opindex mexplicit-relocs
14599 @opindex mno-explicit-relocs
14600 Use (do not use) assembler relocation operators when dealing with symbolic
14601 addresses. The alternative, selected by @option{-mno-explicit-relocs},
14602 is to use assembler macros instead.
14604 @option{-mexplicit-relocs} is the default if GCC was configured
14605 to use an assembler that supports relocation operators.
14607 @item -mcheck-zero-division
14608 @itemx -mno-check-zero-division
14609 @opindex mcheck-zero-division
14610 @opindex mno-check-zero-division
14611 Trap (do not trap) on integer division by zero.
14613 The default is @option{-mcheck-zero-division}.
14615 @item -mdivide-traps
14616 @itemx -mdivide-breaks
14617 @opindex mdivide-traps
14618 @opindex mdivide-breaks
14619 MIPS systems check for division by zero by generating either a
14620 conditional trap or a break instruction. Using traps results in
14621 smaller code, but is only supported on MIPS II and later. Also, some
14622 versions of the Linux kernel have a bug that prevents trap from
14623 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
14624 allow conditional traps on architectures that support them and
14625 @option{-mdivide-breaks} to force the use of breaks.
14627 The default is usually @option{-mdivide-traps}, but this can be
14628 overridden at configure time using @option{--with-divide=breaks}.
14629 Divide-by-zero checks can be completely disabled using
14630 @option{-mno-check-zero-division}.
14635 @opindex mno-memcpy
14636 Force (do not force) the use of @code{memcpy()} for non-trivial block
14637 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
14638 most constant-sized copies.
14641 @itemx -mno-long-calls
14642 @opindex mlong-calls
14643 @opindex mno-long-calls
14644 Disable (do not disable) use of the @code{jal} instruction. Calling
14645 functions using @code{jal} is more efficient but requires the caller
14646 and callee to be in the same 256 megabyte segment.
14648 This option has no effect on abicalls code. The default is
14649 @option{-mno-long-calls}.
14655 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14656 instructions, as provided by the R4650 ISA@.
14659 @itemx -mno-fused-madd
14660 @opindex mfused-madd
14661 @opindex mno-fused-madd
14662 Enable (disable) use of the floating point multiply-accumulate
14663 instructions, when they are available. The default is
14664 @option{-mfused-madd}.
14666 When multiply-accumulate instructions are used, the intermediate
14667 product is calculated to infinite precision and is not subject to
14668 the FCSR Flush to Zero bit. This may be undesirable in some
14673 Tell the MIPS assembler to not run its preprocessor over user
14674 assembler files (with a @samp{.s} suffix) when assembling them.
14677 @itemx -mno-fix-r4000
14678 @opindex mfix-r4000
14679 @opindex mno-fix-r4000
14680 Work around certain R4000 CPU errata:
14683 A double-word or a variable shift may give an incorrect result if executed
14684 immediately after starting an integer division.
14686 A double-word or a variable shift may give an incorrect result if executed
14687 while an integer multiplication is in progress.
14689 An integer division may give an incorrect result if started in a delay slot
14690 of a taken branch or a jump.
14694 @itemx -mno-fix-r4400
14695 @opindex mfix-r4400
14696 @opindex mno-fix-r4400
14697 Work around certain R4400 CPU errata:
14700 A double-word or a variable shift may give an incorrect result if executed
14701 immediately after starting an integer division.
14705 @itemx -mno-fix-r10000
14706 @opindex mfix-r10000
14707 @opindex mno-fix-r10000
14708 Work around certain R10000 errata:
14711 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14712 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14715 This option can only be used if the target architecture supports
14716 branch-likely instructions. @option{-mfix-r10000} is the default when
14717 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14721 @itemx -mno-fix-vr4120
14722 @opindex mfix-vr4120
14723 Work around certain VR4120 errata:
14726 @code{dmultu} does not always produce the correct result.
14728 @code{div} and @code{ddiv} do not always produce the correct result if one
14729 of the operands is negative.
14731 The workarounds for the division errata rely on special functions in
14732 @file{libgcc.a}. At present, these functions are only provided by
14733 the @code{mips64vr*-elf} configurations.
14735 Other VR4120 errata require a nop to be inserted between certain pairs of
14736 instructions. These errata are handled by the assembler, not by GCC itself.
14739 @opindex mfix-vr4130
14740 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14741 workarounds are implemented by the assembler rather than by GCC,
14742 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14743 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14744 instructions are available instead.
14747 @itemx -mno-fix-sb1
14749 Work around certain SB-1 CPU core errata.
14750 (This flag currently works around the SB-1 revision 2
14751 ``F1'' and ``F2'' floating point errata.)
14753 @item -mr10k-cache-barrier=@var{setting}
14754 @opindex mr10k-cache-barrier
14755 Specify whether GCC should insert cache barriers to avoid the
14756 side-effects of speculation on R10K processors.
14758 In common with many processors, the R10K tries to predict the outcome
14759 of a conditional branch and speculatively executes instructions from
14760 the ``taken'' branch. It later aborts these instructions if the
14761 predicted outcome was wrong. However, on the R10K, even aborted
14762 instructions can have side effects.
14764 This problem only affects kernel stores and, depending on the system,
14765 kernel loads. As an example, a speculatively-executed store may load
14766 the target memory into cache and mark the cache line as dirty, even if
14767 the store itself is later aborted. If a DMA operation writes to the
14768 same area of memory before the ``dirty'' line is flushed, the cached
14769 data will overwrite the DMA-ed data. See the R10K processor manual
14770 for a full description, including other potential problems.
14772 One workaround is to insert cache barrier instructions before every memory
14773 access that might be speculatively executed and that might have side
14774 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14775 controls GCC's implementation of this workaround. It assumes that
14776 aborted accesses to any byte in the following regions will not have
14781 the memory occupied by the current function's stack frame;
14784 the memory occupied by an incoming stack argument;
14787 the memory occupied by an object with a link-time-constant address.
14790 It is the kernel's responsibility to ensure that speculative
14791 accesses to these regions are indeed safe.
14793 If the input program contains a function declaration such as:
14799 then the implementation of @code{foo} must allow @code{j foo} and
14800 @code{jal foo} to be executed speculatively. GCC honors this
14801 restriction for functions it compiles itself. It expects non-GCC
14802 functions (such as hand-written assembly code) to do the same.
14804 The option has three forms:
14807 @item -mr10k-cache-barrier=load-store
14808 Insert a cache barrier before a load or store that might be
14809 speculatively executed and that might have side effects even
14812 @item -mr10k-cache-barrier=store
14813 Insert a cache barrier before a store that might be speculatively
14814 executed and that might have side effects even if aborted.
14816 @item -mr10k-cache-barrier=none
14817 Disable the insertion of cache barriers. This is the default setting.
14820 @item -mflush-func=@var{func}
14821 @itemx -mno-flush-func
14822 @opindex mflush-func
14823 Specifies the function to call to flush the I and D caches, or to not
14824 call any such function. If called, the function must take the same
14825 arguments as the common @code{_flush_func()}, that is, the address of the
14826 memory range for which the cache is being flushed, the size of the
14827 memory range, and the number 3 (to flush both caches). The default
14828 depends on the target GCC was configured for, but commonly is either
14829 @samp{_flush_func} or @samp{__cpu_flush}.
14831 @item mbranch-cost=@var{num}
14832 @opindex mbranch-cost
14833 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14834 This cost is only a heuristic and is not guaranteed to produce
14835 consistent results across releases. A zero cost redundantly selects
14836 the default, which is based on the @option{-mtune} setting.
14838 @item -mbranch-likely
14839 @itemx -mno-branch-likely
14840 @opindex mbranch-likely
14841 @opindex mno-branch-likely
14842 Enable or disable use of Branch Likely instructions, regardless of the
14843 default for the selected architecture. By default, Branch Likely
14844 instructions may be generated if they are supported by the selected
14845 architecture. An exception is for the MIPS32 and MIPS64 architectures
14846 and processors which implement those architectures; for those, Branch
14847 Likely instructions will not be generated by default because the MIPS32
14848 and MIPS64 architectures specifically deprecate their use.
14850 @item -mfp-exceptions
14851 @itemx -mno-fp-exceptions
14852 @opindex mfp-exceptions
14853 Specifies whether FP exceptions are enabled. This affects how we schedule
14854 FP instructions for some processors. The default is that FP exceptions are
14857 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14858 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14861 @item -mvr4130-align
14862 @itemx -mno-vr4130-align
14863 @opindex mvr4130-align
14864 The VR4130 pipeline is two-way superscalar, but can only issue two
14865 instructions together if the first one is 8-byte aligned. When this
14866 option is enabled, GCC will align pairs of instructions that it
14867 thinks should execute in parallel.
14869 This option only has an effect when optimizing for the VR4130.
14870 It normally makes code faster, but at the expense of making it bigger.
14871 It is enabled by default at optimization level @option{-O3}.
14876 Enable (disable) generation of @code{synci} instructions on
14877 architectures that support it. The @code{synci} instructions (if
14878 enabled) will be generated when @code{__builtin___clear_cache()} is
14881 This option defaults to @code{-mno-synci}, but the default can be
14882 overridden by configuring with @code{--with-synci}.
14884 When compiling code for single processor systems, it is generally safe
14885 to use @code{synci}. However, on many multi-core (SMP) systems, it
14886 will not invalidate the instruction caches on all cores and may lead
14887 to undefined behavior.
14889 @item -mrelax-pic-calls
14890 @itemx -mno-relax-pic-calls
14891 @opindex mrelax-pic-calls
14892 Try to turn PIC calls that are normally dispatched via register
14893 @code{$25} into direct calls. This is only possible if the linker can
14894 resolve the destination at link-time and if the destination is within
14895 range for a direct call.
14897 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14898 an assembler and a linker that supports the @code{.reloc} assembly
14899 directive and @code{-mexplicit-relocs} is in effect. With
14900 @code{-mno-explicit-relocs}, this optimization can be performed by the
14901 assembler and the linker alone without help from the compiler.
14903 @item -mmcount-ra-address
14904 @itemx -mno-mcount-ra-address
14905 @opindex mmcount-ra-address
14906 @opindex mno-mcount-ra-address
14907 Emit (do not emit) code that allows @code{_mcount} to modify the
14908 calling function's return address. When enabled, this option extends
14909 the usual @code{_mcount} interface with a new @var{ra-address}
14910 parameter, which has type @code{intptr_t *} and is passed in register
14911 @code{$12}. @code{_mcount} can then modify the return address by
14912 doing both of the following:
14915 Returning the new address in register @code{$31}.
14917 Storing the new address in @code{*@var{ra-address}},
14918 if @var{ra-address} is nonnull.
14921 The default is @option{-mno-mcount-ra-address}.
14926 @subsection MMIX Options
14927 @cindex MMIX Options
14929 These options are defined for the MMIX:
14933 @itemx -mno-libfuncs
14935 @opindex mno-libfuncs
14936 Specify that intrinsic library functions are being compiled, passing all
14937 values in registers, no matter the size.
14940 @itemx -mno-epsilon
14942 @opindex mno-epsilon
14943 Generate floating-point comparison instructions that compare with respect
14944 to the @code{rE} epsilon register.
14946 @item -mabi=mmixware
14948 @opindex mabi=mmixware
14950 Generate code that passes function parameters and return values that (in
14951 the called function) are seen as registers @code{$0} and up, as opposed to
14952 the GNU ABI which uses global registers @code{$231} and up.
14954 @item -mzero-extend
14955 @itemx -mno-zero-extend
14956 @opindex mzero-extend
14957 @opindex mno-zero-extend
14958 When reading data from memory in sizes shorter than 64 bits, use (do not
14959 use) zero-extending load instructions by default, rather than
14960 sign-extending ones.
14963 @itemx -mno-knuthdiv
14965 @opindex mno-knuthdiv
14966 Make the result of a division yielding a remainder have the same sign as
14967 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14968 remainder follows the sign of the dividend. Both methods are
14969 arithmetically valid, the latter being almost exclusively used.
14971 @item -mtoplevel-symbols
14972 @itemx -mno-toplevel-symbols
14973 @opindex mtoplevel-symbols
14974 @opindex mno-toplevel-symbols
14975 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14976 code can be used with the @code{PREFIX} assembly directive.
14980 Generate an executable in the ELF format, rather than the default
14981 @samp{mmo} format used by the @command{mmix} simulator.
14983 @item -mbranch-predict
14984 @itemx -mno-branch-predict
14985 @opindex mbranch-predict
14986 @opindex mno-branch-predict
14987 Use (do not use) the probable-branch instructions, when static branch
14988 prediction indicates a probable branch.
14990 @item -mbase-addresses
14991 @itemx -mno-base-addresses
14992 @opindex mbase-addresses
14993 @opindex mno-base-addresses
14994 Generate (do not generate) code that uses @emph{base addresses}. Using a
14995 base address automatically generates a request (handled by the assembler
14996 and the linker) for a constant to be set up in a global register. The
14997 register is used for one or more base address requests within the range 0
14998 to 255 from the value held in the register. The generally leads to short
14999 and fast code, but the number of different data items that can be
15000 addressed is limited. This means that a program that uses lots of static
15001 data may require @option{-mno-base-addresses}.
15003 @item -msingle-exit
15004 @itemx -mno-single-exit
15005 @opindex msingle-exit
15006 @opindex mno-single-exit
15007 Force (do not force) generated code to have a single exit point in each
15011 @node MN10300 Options
15012 @subsection MN10300 Options
15013 @cindex MN10300 options
15015 These @option{-m} options are defined for Matsushita MN10300 architectures:
15020 Generate code to avoid bugs in the multiply instructions for the MN10300
15021 processors. This is the default.
15023 @item -mno-mult-bug
15024 @opindex mno-mult-bug
15025 Do not generate code to avoid bugs in the multiply instructions for the
15026 MN10300 processors.
15030 Generate code which uses features specific to the AM33 processor.
15034 Do not generate code which uses features specific to the AM33 processor. This
15039 Generate code which uses features specific to the AM33/2.0 processor.
15043 Generate code which uses features specific to the AM34 processor.
15045 @item -mtune=@var{cpu-type}
15047 Use the timing characteristics of the indicated CPU type when
15048 scheduling instructions. This does not change the targeted processor
15049 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
15050 @samp{am33-2} or @samp{am34}.
15052 @item -mreturn-pointer-on-d0
15053 @opindex mreturn-pointer-on-d0
15054 When generating a function which returns a pointer, return the pointer
15055 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
15056 only in a0, and attempts to call such functions without a prototype
15057 would result in errors. Note that this option is on by default; use
15058 @option{-mno-return-pointer-on-d0} to disable it.
15062 Do not link in the C run-time initialization object file.
15066 Indicate to the linker that it should perform a relaxation optimization pass
15067 to shorten branches, calls and absolute memory addresses. This option only
15068 has an effect when used on the command line for the final link step.
15070 This option makes symbolic debugging impossible.
15074 Allow the compiler to generate @emph{Long Instruction Word}
15075 instructions if the target is the @samp{AM33} or later. This is the
15076 default. This option defines the preprocessor macro @samp{__LIW__}.
15080 Do not allow the compiler to generate @emph{Long Instruction Word}
15081 instructions. This option defines the preprocessor macro
15086 @node PDP-11 Options
15087 @subsection PDP-11 Options
15088 @cindex PDP-11 Options
15090 These options are defined for the PDP-11:
15095 Use hardware FPP floating point. This is the default. (FIS floating
15096 point on the PDP-11/40 is not supported.)
15099 @opindex msoft-float
15100 Do not use hardware floating point.
15104 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
15108 Return floating-point results in memory. This is the default.
15112 Generate code for a PDP-11/40.
15116 Generate code for a PDP-11/45. This is the default.
15120 Generate code for a PDP-11/10.
15122 @item -mbcopy-builtin
15123 @opindex mbcopy-builtin
15124 Use inline @code{movmemhi} patterns for copying memory. This is the
15129 Do not use inline @code{movmemhi} patterns for copying memory.
15135 Use 16-bit @code{int}. This is the default.
15141 Use 32-bit @code{int}.
15144 @itemx -mno-float32
15146 @opindex mno-float32
15147 Use 64-bit @code{float}. This is the default.
15150 @itemx -mno-float64
15152 @opindex mno-float64
15153 Use 32-bit @code{float}.
15157 Use @code{abshi2} pattern. This is the default.
15161 Do not use @code{abshi2} pattern.
15163 @item -mbranch-expensive
15164 @opindex mbranch-expensive
15165 Pretend that branches are expensive. This is for experimenting with
15166 code generation only.
15168 @item -mbranch-cheap
15169 @opindex mbranch-cheap
15170 Do not pretend that branches are expensive. This is the default.
15174 Use Unix assembler syntax. This is the default when configured for
15175 @samp{pdp11-*-bsd}.
15179 Use DEC assembler syntax. This is the default when configured for any
15180 PDP-11 target other than @samp{pdp11-*-bsd}.
15183 @node picoChip Options
15184 @subsection picoChip Options
15185 @cindex picoChip options
15187 These @samp{-m} options are defined for picoChip implementations:
15191 @item -mae=@var{ae_type}
15193 Set the instruction set, register set, and instruction scheduling
15194 parameters for array element type @var{ae_type}. Supported values
15195 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
15197 @option{-mae=ANY} selects a completely generic AE type. Code
15198 generated with this option will run on any of the other AE types. The
15199 code will not be as efficient as it would be if compiled for a specific
15200 AE type, and some types of operation (e.g., multiplication) will not
15201 work properly on all types of AE.
15203 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
15204 for compiled code, and is the default.
15206 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
15207 option may suffer from poor performance of byte (char) manipulation,
15208 since the DSP AE does not provide hardware support for byte load/stores.
15210 @item -msymbol-as-address
15211 Enable the compiler to directly use a symbol name as an address in a
15212 load/store instruction, without first loading it into a
15213 register. Typically, the use of this option will generate larger
15214 programs, which run faster than when the option isn't used. However, the
15215 results vary from program to program, so it is left as a user option,
15216 rather than being permanently enabled.
15218 @item -mno-inefficient-warnings
15219 Disables warnings about the generation of inefficient code. These
15220 warnings can be generated, for example, when compiling code which
15221 performs byte-level memory operations on the MAC AE type. The MAC AE has
15222 no hardware support for byte-level memory operations, so all byte
15223 load/stores must be synthesized from word load/store operations. This is
15224 inefficient and a warning will be generated indicating to the programmer
15225 that they should rewrite the code to avoid byte operations, or to target
15226 an AE type which has the necessary hardware support. This option enables
15227 the warning to be turned off.
15231 @node PowerPC Options
15232 @subsection PowerPC Options
15233 @cindex PowerPC options
15235 These are listed under @xref{RS/6000 and PowerPC Options}.
15237 @node RS/6000 and PowerPC Options
15238 @subsection IBM RS/6000 and PowerPC Options
15239 @cindex RS/6000 and PowerPC Options
15240 @cindex IBM RS/6000 and PowerPC Options
15242 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
15249 @itemx -mno-powerpc
15250 @itemx -mpowerpc-gpopt
15251 @itemx -mno-powerpc-gpopt
15252 @itemx -mpowerpc-gfxopt
15253 @itemx -mno-powerpc-gfxopt
15256 @itemx -mno-powerpc64
15260 @itemx -mno-popcntb
15262 @itemx -mno-popcntd
15271 @itemx -mno-hard-dfp
15275 @opindex mno-power2
15277 @opindex mno-powerpc
15278 @opindex mpowerpc-gpopt
15279 @opindex mno-powerpc-gpopt
15280 @opindex mpowerpc-gfxopt
15281 @opindex mno-powerpc-gfxopt
15282 @opindex mpowerpc64
15283 @opindex mno-powerpc64
15287 @opindex mno-popcntb
15289 @opindex mno-popcntd
15295 @opindex mno-mfpgpr
15297 @opindex mno-hard-dfp
15298 GCC supports two related instruction set architectures for the
15299 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
15300 instructions supported by the @samp{rios} chip set used in the original
15301 RS/6000 systems and the @dfn{PowerPC} instruction set is the
15302 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
15303 the IBM 4xx, 6xx, and follow-on microprocessors.
15305 Neither architecture is a subset of the other. However there is a
15306 large common subset of instructions supported by both. An MQ
15307 register is included in processors supporting the POWER architecture.
15309 You use these options to specify which instructions are available on the
15310 processor you are using. The default value of these options is
15311 determined when configuring GCC@. Specifying the
15312 @option{-mcpu=@var{cpu_type}} overrides the specification of these
15313 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
15314 rather than the options listed above.
15316 The @option{-mpower} option allows GCC to generate instructions that
15317 are found only in the POWER architecture and to use the MQ register.
15318 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
15319 to generate instructions that are present in the POWER2 architecture but
15320 not the original POWER architecture.
15322 The @option{-mpowerpc} option allows GCC to generate instructions that
15323 are found only in the 32-bit subset of the PowerPC architecture.
15324 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
15325 GCC to use the optional PowerPC architecture instructions in the
15326 General Purpose group, including floating-point square root. Specifying
15327 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
15328 use the optional PowerPC architecture instructions in the Graphics
15329 group, including floating-point select.
15331 The @option{-mmfcrf} option allows GCC to generate the move from
15332 condition register field instruction implemented on the POWER4
15333 processor and other processors that support the PowerPC V2.01
15335 The @option{-mpopcntb} option allows GCC to generate the popcount and
15336 double precision FP reciprocal estimate instruction implemented on the
15337 POWER5 processor and other processors that support the PowerPC V2.02
15339 The @option{-mpopcntd} option allows GCC to generate the popcount
15340 instruction implemented on the POWER7 processor and other processors
15341 that support the PowerPC V2.06 architecture.
15342 The @option{-mfprnd} option allows GCC to generate the FP round to
15343 integer instructions implemented on the POWER5+ processor and other
15344 processors that support the PowerPC V2.03 architecture.
15345 The @option{-mcmpb} option allows GCC to generate the compare bytes
15346 instruction implemented on the POWER6 processor and other processors
15347 that support the PowerPC V2.05 architecture.
15348 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
15349 general purpose register instructions implemented on the POWER6X
15350 processor and other processors that support the extended PowerPC V2.05
15352 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
15353 point instructions implemented on some POWER processors.
15355 The @option{-mpowerpc64} option allows GCC to generate the additional
15356 64-bit instructions that are found in the full PowerPC64 architecture
15357 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
15358 @option{-mno-powerpc64}.
15360 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
15361 will use only the instructions in the common subset of both
15362 architectures plus some special AIX common-mode calls, and will not use
15363 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
15364 permits GCC to use any instruction from either architecture and to
15365 allow use of the MQ register; specify this for the Motorola MPC601.
15367 @item -mnew-mnemonics
15368 @itemx -mold-mnemonics
15369 @opindex mnew-mnemonics
15370 @opindex mold-mnemonics
15371 Select which mnemonics to use in the generated assembler code. With
15372 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
15373 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
15374 assembler mnemonics defined for the POWER architecture. Instructions
15375 defined in only one architecture have only one mnemonic; GCC uses that
15376 mnemonic irrespective of which of these options is specified.
15378 GCC defaults to the mnemonics appropriate for the architecture in
15379 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
15380 value of these option. Unless you are building a cross-compiler, you
15381 should normally not specify either @option{-mnew-mnemonics} or
15382 @option{-mold-mnemonics}, but should instead accept the default.
15384 @item -mcpu=@var{cpu_type}
15386 Set architecture type, register usage, choice of mnemonics, and
15387 instruction scheduling parameters for machine type @var{cpu_type}.
15388 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
15389 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
15390 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
15391 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
15392 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
15393 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
15394 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
15395 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
15396 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
15397 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
15398 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
15400 @option{-mcpu=common} selects a completely generic processor. Code
15401 generated under this option will run on any POWER or PowerPC processor.
15402 GCC will use only the instructions in the common subset of both
15403 architectures, and will not use the MQ register. GCC assumes a generic
15404 processor model for scheduling purposes.
15406 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
15407 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
15408 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
15409 types, with an appropriate, generic processor model assumed for
15410 scheduling purposes.
15412 The other options specify a specific processor. Code generated under
15413 those options will run best on that processor, and may not run at all on
15416 The @option{-mcpu} options automatically enable or disable the
15419 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
15420 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
15421 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
15422 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
15424 The particular options set for any particular CPU will vary between
15425 compiler versions, depending on what setting seems to produce optimal
15426 code for that CPU; it doesn't necessarily reflect the actual hardware's
15427 capabilities. If you wish to set an individual option to a particular
15428 value, you may specify it after the @option{-mcpu} option, like
15429 @samp{-mcpu=970 -mno-altivec}.
15431 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
15432 not enabled or disabled by the @option{-mcpu} option at present because
15433 AIX does not have full support for these options. You may still
15434 enable or disable them individually if you're sure it'll work in your
15437 @item -mtune=@var{cpu_type}
15439 Set the instruction scheduling parameters for machine type
15440 @var{cpu_type}, but do not set the architecture type, register usage, or
15441 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
15442 values for @var{cpu_type} are used for @option{-mtune} as for
15443 @option{-mcpu}. If both are specified, the code generated will use the
15444 architecture, registers, and mnemonics set by @option{-mcpu}, but the
15445 scheduling parameters set by @option{-mtune}.
15447 @item -mcmodel=small
15448 @opindex mcmodel=small
15449 Generate PowerPC64 code for the small model: The TOC is limited to
15452 @item -mcmodel=medium
15453 @opindex mcmodel=medium
15454 Generate PowerPC64 code for the medium model: The TOC and other static
15455 data may be up to a total of 4G in size.
15457 @item -mcmodel=large
15458 @opindex mcmodel=large
15459 Generate PowerPC64 code for the large model: The TOC may be up to 4G
15460 in size. Other data and code is only limited by the 64-bit address
15464 @itemx -mno-altivec
15466 @opindex mno-altivec
15467 Generate code that uses (does not use) AltiVec instructions, and also
15468 enable the use of built-in functions that allow more direct access to
15469 the AltiVec instruction set. You may also need to set
15470 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
15476 @opindex mno-vrsave
15477 Generate VRSAVE instructions when generating AltiVec code.
15479 @item -mgen-cell-microcode
15480 @opindex mgen-cell-microcode
15481 Generate Cell microcode instructions
15483 @item -mwarn-cell-microcode
15484 @opindex mwarn-cell-microcode
15485 Warning when a Cell microcode instruction is going to emitted. An example
15486 of a Cell microcode instruction is a variable shift.
15489 @opindex msecure-plt
15490 Generate code that allows ld and ld.so to build executables and shared
15491 libraries with non-exec .plt and .got sections. This is a PowerPC
15492 32-bit SYSV ABI option.
15496 Generate code that uses a BSS .plt section that ld.so fills in, and
15497 requires .plt and .got sections that are both writable and executable.
15498 This is a PowerPC 32-bit SYSV ABI option.
15504 This switch enables or disables the generation of ISEL instructions.
15506 @item -misel=@var{yes/no}
15507 This switch has been deprecated. Use @option{-misel} and
15508 @option{-mno-isel} instead.
15514 This switch enables or disables the generation of SPE simd
15520 @opindex mno-paired
15521 This switch enables or disables the generation of PAIRED simd
15524 @item -mspe=@var{yes/no}
15525 This option has been deprecated. Use @option{-mspe} and
15526 @option{-mno-spe} instead.
15532 Generate code that uses (does not use) vector/scalar (VSX)
15533 instructions, and also enable the use of built-in functions that allow
15534 more direct access to the VSX instruction set.
15536 @item -mfloat-gprs=@var{yes/single/double/no}
15537 @itemx -mfloat-gprs
15538 @opindex mfloat-gprs
15539 This switch enables or disables the generation of floating point
15540 operations on the general purpose registers for architectures that
15543 The argument @var{yes} or @var{single} enables the use of
15544 single-precision floating point operations.
15546 The argument @var{double} enables the use of single and
15547 double-precision floating point operations.
15549 The argument @var{no} disables floating point operations on the
15550 general purpose registers.
15552 This option is currently only available on the MPC854x.
15558 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
15559 targets (including GNU/Linux). The 32-bit environment sets int, long
15560 and pointer to 32 bits and generates code that runs on any PowerPC
15561 variant. The 64-bit environment sets int to 32 bits and long and
15562 pointer to 64 bits, and generates code for PowerPC64, as for
15563 @option{-mpowerpc64}.
15566 @itemx -mno-fp-in-toc
15567 @itemx -mno-sum-in-toc
15568 @itemx -mminimal-toc
15570 @opindex mno-fp-in-toc
15571 @opindex mno-sum-in-toc
15572 @opindex mminimal-toc
15573 Modify generation of the TOC (Table Of Contents), which is created for
15574 every executable file. The @option{-mfull-toc} option is selected by
15575 default. In that case, GCC will allocate at least one TOC entry for
15576 each unique non-automatic variable reference in your program. GCC
15577 will also place floating-point constants in the TOC@. However, only
15578 16,384 entries are available in the TOC@.
15580 If you receive a linker error message that saying you have overflowed
15581 the available TOC space, you can reduce the amount of TOC space used
15582 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
15583 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
15584 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
15585 generate code to calculate the sum of an address and a constant at
15586 run-time instead of putting that sum into the TOC@. You may specify one
15587 or both of these options. Each causes GCC to produce very slightly
15588 slower and larger code at the expense of conserving TOC space.
15590 If you still run out of space in the TOC even when you specify both of
15591 these options, specify @option{-mminimal-toc} instead. This option causes
15592 GCC to make only one TOC entry for every file. When you specify this
15593 option, GCC will produce code that is slower and larger but which
15594 uses extremely little TOC space. You may wish to use this option
15595 only on files that contain less frequently executed code.
15601 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
15602 @code{long} type, and the infrastructure needed to support them.
15603 Specifying @option{-maix64} implies @option{-mpowerpc64} and
15604 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
15605 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
15608 @itemx -mno-xl-compat
15609 @opindex mxl-compat
15610 @opindex mno-xl-compat
15611 Produce code that conforms more closely to IBM XL compiler semantics
15612 when using AIX-compatible ABI@. Pass floating-point arguments to
15613 prototyped functions beyond the register save area (RSA) on the stack
15614 in addition to argument FPRs. Do not assume that most significant
15615 double in 128-bit long double value is properly rounded when comparing
15616 values and converting to double. Use XL symbol names for long double
15619 The AIX calling convention was extended but not initially documented to
15620 handle an obscure K&R C case of calling a function that takes the
15621 address of its arguments with fewer arguments than declared. IBM XL
15622 compilers access floating point arguments which do not fit in the
15623 RSA from the stack when a subroutine is compiled without
15624 optimization. Because always storing floating-point arguments on the
15625 stack is inefficient and rarely needed, this option is not enabled by
15626 default and only is necessary when calling subroutines compiled by IBM
15627 XL compilers without optimization.
15631 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
15632 application written to use message passing with special startup code to
15633 enable the application to run. The system must have PE installed in the
15634 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
15635 must be overridden with the @option{-specs=} option to specify the
15636 appropriate directory location. The Parallel Environment does not
15637 support threads, so the @option{-mpe} option and the @option{-pthread}
15638 option are incompatible.
15640 @item -malign-natural
15641 @itemx -malign-power
15642 @opindex malign-natural
15643 @opindex malign-power
15644 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
15645 @option{-malign-natural} overrides the ABI-defined alignment of larger
15646 types, such as floating-point doubles, on their natural size-based boundary.
15647 The option @option{-malign-power} instructs GCC to follow the ABI-specified
15648 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
15650 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
15654 @itemx -mhard-float
15655 @opindex msoft-float
15656 @opindex mhard-float
15657 Generate code that does not use (uses) the floating-point register set.
15658 Software floating point emulation is provided if you use the
15659 @option{-msoft-float} option, and pass the option to GCC when linking.
15661 @item -msingle-float
15662 @itemx -mdouble-float
15663 @opindex msingle-float
15664 @opindex mdouble-float
15665 Generate code for single or double-precision floating point operations.
15666 @option{-mdouble-float} implies @option{-msingle-float}.
15669 @opindex msimple-fpu
15670 Do not generate sqrt and div instructions for hardware floating point unit.
15674 Specify type of floating point unit. Valid values are @var{sp_lite}
15675 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15676 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15677 and @var{dp_full} (equivalent to -mdouble-float).
15680 @opindex mxilinx-fpu
15681 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15684 @itemx -mno-multiple
15686 @opindex mno-multiple
15687 Generate code that uses (does not use) the load multiple word
15688 instructions and the store multiple word instructions. These
15689 instructions are generated by default on POWER systems, and not
15690 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15691 endian PowerPC systems, since those instructions do not work when the
15692 processor is in little endian mode. The exceptions are PPC740 and
15693 PPC750 which permit the instructions usage in little endian mode.
15698 @opindex mno-string
15699 Generate code that uses (does not use) the load string instructions
15700 and the store string word instructions to save multiple registers and
15701 do small block moves. These instructions are generated by default on
15702 POWER systems, and not generated on PowerPC systems. Do not use
15703 @option{-mstring} on little endian PowerPC systems, since those
15704 instructions do not work when the processor is in little endian mode.
15705 The exceptions are PPC740 and PPC750 which permit the instructions
15706 usage in little endian mode.
15711 @opindex mno-update
15712 Generate code that uses (does not use) the load or store instructions
15713 that update the base register to the address of the calculated memory
15714 location. These instructions are generated by default. If you use
15715 @option{-mno-update}, there is a small window between the time that the
15716 stack pointer is updated and the address of the previous frame is
15717 stored, which means code that walks the stack frame across interrupts or
15718 signals may get corrupted data.
15720 @item -mavoid-indexed-addresses
15721 @itemx -mno-avoid-indexed-addresses
15722 @opindex mavoid-indexed-addresses
15723 @opindex mno-avoid-indexed-addresses
15724 Generate code that tries to avoid (not avoid) the use of indexed load
15725 or store instructions. These instructions can incur a performance
15726 penalty on Power6 processors in certain situations, such as when
15727 stepping through large arrays that cross a 16M boundary. This option
15728 is enabled by default when targetting Power6 and disabled otherwise.
15731 @itemx -mno-fused-madd
15732 @opindex mfused-madd
15733 @opindex mno-fused-madd
15734 Generate code that uses (does not use) the floating point multiply and
15735 accumulate instructions. These instructions are generated by default
15736 if hardware floating point is used. The machine dependent
15737 @option{-mfused-madd} option is now mapped to the machine independent
15738 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
15739 mapped to @option{-ffp-contract=off}.
15745 Generate code that uses (does not use) the half-word multiply and
15746 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15747 These instructions are generated by default when targetting those
15754 Generate code that uses (does not use) the string-search @samp{dlmzb}
15755 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15756 generated by default when targetting those processors.
15758 @item -mno-bit-align
15760 @opindex mno-bit-align
15761 @opindex mbit-align
15762 On System V.4 and embedded PowerPC systems do not (do) force structures
15763 and unions that contain bit-fields to be aligned to the base type of the
15766 For example, by default a structure containing nothing but 8
15767 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15768 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15769 the structure would be aligned to a 1 byte boundary and be one byte in
15772 @item -mno-strict-align
15773 @itemx -mstrict-align
15774 @opindex mno-strict-align
15775 @opindex mstrict-align
15776 On System V.4 and embedded PowerPC systems do not (do) assume that
15777 unaligned memory references will be handled by the system.
15779 @item -mrelocatable
15780 @itemx -mno-relocatable
15781 @opindex mrelocatable
15782 @opindex mno-relocatable
15783 Generate code that allows (does not allow) a static executable to be
15784 relocated to a different address at runtime. A simple embedded
15785 PowerPC system loader should relocate the entire contents of
15786 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
15787 a table of 32-bit addresses generated by this option. For this to
15788 work, all objects linked together must be compiled with
15789 @option{-mrelocatable} or @option{-mrelocatable-lib}.
15790 @option{-mrelocatable} code aligns the stack to an 8 byte boundary.
15792 @item -mrelocatable-lib
15793 @itemx -mno-relocatable-lib
15794 @opindex mrelocatable-lib
15795 @opindex mno-relocatable-lib
15796 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
15797 @code{.fixup} section to allow static executables to be relocated at
15798 runtime, but @option{-mrelocatable-lib} does not use the smaller stack
15799 alignment of @option{-mrelocatable}. Objects compiled with
15800 @option{-mrelocatable-lib} may be linked with objects compiled with
15801 any combination of the @option{-mrelocatable} options.
15807 On System V.4 and embedded PowerPC systems do not (do) assume that
15808 register 2 contains a pointer to a global area pointing to the addresses
15809 used in the program.
15812 @itemx -mlittle-endian
15814 @opindex mlittle-endian
15815 On System V.4 and embedded PowerPC systems compile code for the
15816 processor in little endian mode. The @option{-mlittle-endian} option is
15817 the same as @option{-mlittle}.
15820 @itemx -mbig-endian
15822 @opindex mbig-endian
15823 On System V.4 and embedded PowerPC systems compile code for the
15824 processor in big endian mode. The @option{-mbig-endian} option is
15825 the same as @option{-mbig}.
15827 @item -mdynamic-no-pic
15828 @opindex mdynamic-no-pic
15829 On Darwin and Mac OS X systems, compile code so that it is not
15830 relocatable, but that its external references are relocatable. The
15831 resulting code is suitable for applications, but not shared
15834 @item -msingle-pic-base
15835 @opindex msingle-pic-base
15836 Treat the register used for PIC addressing as read-only, rather than
15837 loading it in the prologue for each function. The run-time system is
15838 responsible for initializing this register with an appropriate value
15839 before execution begins.
15841 @item -mprioritize-restricted-insns=@var{priority}
15842 @opindex mprioritize-restricted-insns
15843 This option controls the priority that is assigned to
15844 dispatch-slot restricted instructions during the second scheduling
15845 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15846 @var{no/highest/second-highest} priority to dispatch slot restricted
15849 @item -msched-costly-dep=@var{dependence_type}
15850 @opindex msched-costly-dep
15851 This option controls which dependences are considered costly
15852 by the target during instruction scheduling. The argument
15853 @var{dependence_type} takes one of the following values:
15854 @var{no}: no dependence is costly,
15855 @var{all}: all dependences are costly,
15856 @var{true_store_to_load}: a true dependence from store to load is costly,
15857 @var{store_to_load}: any dependence from store to load is costly,
15858 @var{number}: any dependence which latency >= @var{number} is costly.
15860 @item -minsert-sched-nops=@var{scheme}
15861 @opindex minsert-sched-nops
15862 This option controls which nop insertion scheme will be used during
15863 the second scheduling pass. The argument @var{scheme} takes one of the
15865 @var{no}: Don't insert nops.
15866 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15867 according to the scheduler's grouping.
15868 @var{regroup_exact}: Insert nops to force costly dependent insns into
15869 separate groups. Insert exactly as many nops as needed to force an insn
15870 to a new group, according to the estimated processor grouping.
15871 @var{number}: Insert nops to force costly dependent insns into
15872 separate groups. Insert @var{number} nops to force an insn to a new group.
15875 @opindex mcall-sysv
15876 On System V.4 and embedded PowerPC systems compile code using calling
15877 conventions that adheres to the March 1995 draft of the System V
15878 Application Binary Interface, PowerPC processor supplement. This is the
15879 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15881 @item -mcall-sysv-eabi
15883 @opindex mcall-sysv-eabi
15884 @opindex mcall-eabi
15885 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15887 @item -mcall-sysv-noeabi
15888 @opindex mcall-sysv-noeabi
15889 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15891 @item -mcall-aixdesc
15893 On System V.4 and embedded PowerPC systems compile code for the AIX
15897 @opindex mcall-linux
15898 On System V.4 and embedded PowerPC systems compile code for the
15899 Linux-based GNU system.
15903 On System V.4 and embedded PowerPC systems compile code for the
15904 Hurd-based GNU system.
15906 @item -mcall-freebsd
15907 @opindex mcall-freebsd
15908 On System V.4 and embedded PowerPC systems compile code for the
15909 FreeBSD operating system.
15911 @item -mcall-netbsd
15912 @opindex mcall-netbsd
15913 On System V.4 and embedded PowerPC systems compile code for the
15914 NetBSD operating system.
15916 @item -mcall-openbsd
15917 @opindex mcall-netbsd
15918 On System V.4 and embedded PowerPC systems compile code for the
15919 OpenBSD operating system.
15921 @item -maix-struct-return
15922 @opindex maix-struct-return
15923 Return all structures in memory (as specified by the AIX ABI)@.
15925 @item -msvr4-struct-return
15926 @opindex msvr4-struct-return
15927 Return structures smaller than 8 bytes in registers (as specified by the
15930 @item -mabi=@var{abi-type}
15932 Extend the current ABI with a particular extension, or remove such extension.
15933 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15934 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15938 Extend the current ABI with SPE ABI extensions. This does not change
15939 the default ABI, instead it adds the SPE ABI extensions to the current
15943 @opindex mabi=no-spe
15944 Disable Booke SPE ABI extensions for the current ABI@.
15946 @item -mabi=ibmlongdouble
15947 @opindex mabi=ibmlongdouble
15948 Change the current ABI to use IBM extended precision long double.
15949 This is a PowerPC 32-bit SYSV ABI option.
15951 @item -mabi=ieeelongdouble
15952 @opindex mabi=ieeelongdouble
15953 Change the current ABI to use IEEE extended precision long double.
15954 This is a PowerPC 32-bit Linux ABI option.
15957 @itemx -mno-prototype
15958 @opindex mprototype
15959 @opindex mno-prototype
15960 On System V.4 and embedded PowerPC systems assume that all calls to
15961 variable argument functions are properly prototyped. Otherwise, the
15962 compiler must insert an instruction before every non prototyped call to
15963 set or clear bit 6 of the condition code register (@var{CR}) to
15964 indicate whether floating point values were passed in the floating point
15965 registers in case the function takes a variable arguments. With
15966 @option{-mprototype}, only calls to prototyped variable argument functions
15967 will set or clear the bit.
15971 On embedded PowerPC systems, assume that the startup module is called
15972 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15973 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15978 On embedded PowerPC systems, assume that the startup module is called
15979 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15984 On embedded PowerPC systems, assume that the startup module is called
15985 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15988 @item -myellowknife
15989 @opindex myellowknife
15990 On embedded PowerPC systems, assume that the startup module is called
15991 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15996 On System V.4 and embedded PowerPC systems, specify that you are
15997 compiling for a VxWorks system.
16001 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
16002 header to indicate that @samp{eabi} extended relocations are used.
16008 On System V.4 and embedded PowerPC systems do (do not) adhere to the
16009 Embedded Applications Binary Interface (eabi) which is a set of
16010 modifications to the System V.4 specifications. Selecting @option{-meabi}
16011 means that the stack is aligned to an 8 byte boundary, a function
16012 @code{__eabi} is called to from @code{main} to set up the eabi
16013 environment, and the @option{-msdata} option can use both @code{r2} and
16014 @code{r13} to point to two separate small data areas. Selecting
16015 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
16016 do not call an initialization function from @code{main}, and the
16017 @option{-msdata} option will only use @code{r13} to point to a single
16018 small data area. The @option{-meabi} option is on by default if you
16019 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
16022 @opindex msdata=eabi
16023 On System V.4 and embedded PowerPC systems, put small initialized
16024 @code{const} global and static data in the @samp{.sdata2} section, which
16025 is pointed to by register @code{r2}. Put small initialized
16026 non-@code{const} global and static data in the @samp{.sdata} section,
16027 which is pointed to by register @code{r13}. Put small uninitialized
16028 global and static data in the @samp{.sbss} section, which is adjacent to
16029 the @samp{.sdata} section. The @option{-msdata=eabi} option is
16030 incompatible with the @option{-mrelocatable} option. The
16031 @option{-msdata=eabi} option also sets the @option{-memb} option.
16034 @opindex msdata=sysv
16035 On System V.4 and embedded PowerPC systems, put small global and static
16036 data in the @samp{.sdata} section, which is pointed to by register
16037 @code{r13}. Put small uninitialized global and static data in the
16038 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
16039 The @option{-msdata=sysv} option is incompatible with the
16040 @option{-mrelocatable} option.
16042 @item -msdata=default
16044 @opindex msdata=default
16046 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
16047 compile code the same as @option{-msdata=eabi}, otherwise compile code the
16048 same as @option{-msdata=sysv}.
16051 @opindex msdata=data
16052 On System V.4 and embedded PowerPC systems, put small global
16053 data in the @samp{.sdata} section. Put small uninitialized global
16054 data in the @samp{.sbss} section. Do not use register @code{r13}
16055 to address small data however. This is the default behavior unless
16056 other @option{-msdata} options are used.
16060 @opindex msdata=none
16062 On embedded PowerPC systems, put all initialized global and static data
16063 in the @samp{.data} section, and all uninitialized data in the
16064 @samp{.bss} section.
16066 @item -mblock-move-inline-limit=@var{num}
16067 @opindex mblock-move-inline-limit
16068 Inline all block moves (such as calls to @code{memcpy} or structure
16069 copies) less than or equal to @var{num} bytes. The minimum value for
16070 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
16071 targets. The default value is target-specific.
16075 @cindex smaller data references (PowerPC)
16076 @cindex .sdata/.sdata2 references (PowerPC)
16077 On embedded PowerPC systems, put global and static items less than or
16078 equal to @var{num} bytes into the small data or bss sections instead of
16079 the normal data or bss section. By default, @var{num} is 8. The
16080 @option{-G @var{num}} switch is also passed to the linker.
16081 All modules should be compiled with the same @option{-G @var{num}} value.
16084 @itemx -mno-regnames
16086 @opindex mno-regnames
16087 On System V.4 and embedded PowerPC systems do (do not) emit register
16088 names in the assembly language output using symbolic forms.
16091 @itemx -mno-longcall
16093 @opindex mno-longcall
16094 By default assume that all calls are far away so that a longer more
16095 expensive calling sequence is required. This is required for calls
16096 further than 32 megabytes (33,554,432 bytes) from the current location.
16097 A short call will be generated if the compiler knows
16098 the call cannot be that far away. This setting can be overridden by
16099 the @code{shortcall} function attribute, or by @code{#pragma
16102 Some linkers are capable of detecting out-of-range calls and generating
16103 glue code on the fly. On these systems, long calls are unnecessary and
16104 generate slower code. As of this writing, the AIX linker can do this,
16105 as can the GNU linker for PowerPC/64. It is planned to add this feature
16106 to the GNU linker for 32-bit PowerPC systems as well.
16108 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
16109 callee, L42'', plus a ``branch island'' (glue code). The two target
16110 addresses represent the callee and the ``branch island''. The
16111 Darwin/PPC linker will prefer the first address and generate a ``bl
16112 callee'' if the PPC ``bl'' instruction will reach the callee directly;
16113 otherwise, the linker will generate ``bl L42'' to call the ``branch
16114 island''. The ``branch island'' is appended to the body of the
16115 calling function; it computes the full 32-bit address of the callee
16118 On Mach-O (Darwin) systems, this option directs the compiler emit to
16119 the glue for every direct call, and the Darwin linker decides whether
16120 to use or discard it.
16122 In the future, we may cause GCC to ignore all longcall specifications
16123 when the linker is known to generate glue.
16125 @item -mtls-markers
16126 @itemx -mno-tls-markers
16127 @opindex mtls-markers
16128 @opindex mno-tls-markers
16129 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
16130 specifying the function argument. The relocation allows ld to
16131 reliably associate function call with argument setup instructions for
16132 TLS optimization, which in turn allows gcc to better schedule the
16137 Adds support for multithreading with the @dfn{pthreads} library.
16138 This option sets flags for both the preprocessor and linker.
16143 This option will enable GCC to use the reciprocal estimate and
16144 reciprocal square root estimate instructions with additional
16145 Newton-Raphson steps to increase precision instead of doing a divide or
16146 square root and divide for floating point arguments. You should use
16147 the @option{-ffast-math} option when using @option{-mrecip} (or at
16148 least @option{-funsafe-math-optimizations},
16149 @option{-finite-math-only}, @option{-freciprocal-math} and
16150 @option{-fno-trapping-math}). Note that while the throughput of the
16151 sequence is generally higher than the throughput of the non-reciprocal
16152 instruction, the precision of the sequence can be decreased by up to 2
16153 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
16156 @item -mrecip=@var{opt}
16157 @opindex mrecip=opt
16158 This option allows to control which reciprocal estimate instructions
16159 may be used. @var{opt} is a comma separated list of options, that may
16160 be preceded by a @code{!} to invert the option:
16161 @code{all}: enable all estimate instructions,
16162 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
16163 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
16164 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
16165 @code{divf}: enable the single precision reciprocal approximation instructions;
16166 @code{divd}: enable the double precision reciprocal approximation instructions;
16167 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
16168 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
16169 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
16171 So for example, @option{-mrecip=all,!rsqrtd} would enable the
16172 all of the reciprocal estimate instructions, except for the
16173 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
16174 which handle the double precision reciprocal square root calculations.
16176 @item -mrecip-precision
16177 @itemx -mno-recip-precision
16178 @opindex mrecip-precision
16179 Assume (do not assume) that the reciprocal estimate instructions
16180 provide higher precision estimates than is mandated by the powerpc
16181 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
16182 automatically selects @option{-mrecip-precision}. The double
16183 precision square root estimate instructions are not generated by
16184 default on low precision machines, since they do not provide an
16185 estimate that converges after three steps.
16187 @item -mveclibabi=@var{type}
16188 @opindex mveclibabi
16189 Specifies the ABI type to use for vectorizing intrinsics using an
16190 external library. The only type supported at present is @code{mass},
16191 which specifies to use IBM's Mathematical Acceleration Subsystem
16192 (MASS) libraries for vectorizing intrinsics using external libraries.
16193 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
16194 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
16195 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
16196 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
16197 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
16198 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
16199 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
16200 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
16201 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
16202 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
16203 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
16204 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
16205 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
16206 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
16207 for power7. Both @option{-ftree-vectorize} and
16208 @option{-funsafe-math-optimizations} have to be enabled. The MASS
16209 libraries will have to be specified at link time.
16214 Generate (do not generate) the @code{friz} instruction when the
16215 @option{-funsafe-math-optimizations} option is used to optimize
16216 rounding a floating point value to 64-bit integer and back to floating
16217 point. The @code{friz} instruction does not return the same value if
16218 the floating point number is too large to fit in an integer.
16222 @subsection RX Options
16225 These command line options are defined for RX targets:
16228 @item -m64bit-doubles
16229 @itemx -m32bit-doubles
16230 @opindex m64bit-doubles
16231 @opindex m32bit-doubles
16232 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
16233 or 32-bits (@option{-m32bit-doubles}) in size. The default is
16234 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
16235 works on 32-bit values, which is why the default is
16236 @option{-m32bit-doubles}.
16242 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
16243 floating point hardware. The default is enabled for the @var{RX600}
16244 series and disabled for the @var{RX200} series.
16246 Floating point instructions will only be generated for 32-bit floating
16247 point values however, so if the @option{-m64bit-doubles} option is in
16248 use then the FPU hardware will not be used for doubles.
16250 @emph{Note} If the @option{-fpu} option is enabled then
16251 @option{-funsafe-math-optimizations} is also enabled automatically.
16252 This is because the RX FPU instructions are themselves unsafe.
16254 @item -mcpu=@var{name}
16256 Selects the type of RX CPU to be targeted. Currently three types are
16257 supported, the generic @var{RX600} and @var{RX200} series hardware and
16258 the specific @var{RX610} CPU. The default is @var{RX600}.
16260 The only difference between @var{RX600} and @var{RX610} is that the
16261 @var{RX610} does not support the @code{MVTIPL} instruction.
16263 The @var{RX200} series does not have a hardware floating point unit
16264 and so @option{-nofpu} is enabled by default when this type is
16267 @item -mbig-endian-data
16268 @itemx -mlittle-endian-data
16269 @opindex mbig-endian-data
16270 @opindex mlittle-endian-data
16271 Store data (but not code) in the big-endian format. The default is
16272 @option{-mlittle-endian-data}, i.e.@: to store data in the little endian
16275 @item -msmall-data-limit=@var{N}
16276 @opindex msmall-data-limit
16277 Specifies the maximum size in bytes of global and static variables
16278 which can be placed into the small data area. Using the small data
16279 area can lead to smaller and faster code, but the size of area is
16280 limited and it is up to the programmer to ensure that the area does
16281 not overflow. Also when the small data area is used one of the RX's
16282 registers (@code{r13}) is reserved for use pointing to this area, so
16283 it is no longer available for use by the compiler. This could result
16284 in slower and/or larger code if variables which once could have been
16285 held in @code{r13} are now pushed onto the stack.
16287 Note, common variables (variables which have not been initialised) and
16288 constants are not placed into the small data area as they are assigned
16289 to other sections in the output executable.
16291 The default value is zero, which disables this feature. Note, this
16292 feature is not enabled by default with higher optimization levels
16293 (@option{-O2} etc) because of the potentially detrimental effects of
16294 reserving register @code{r13}. It is up to the programmer to
16295 experiment and discover whether this feature is of benefit to their
16302 Use the simulator runtime. The default is to use the libgloss board
16305 @item -mas100-syntax
16306 @itemx -mno-as100-syntax
16307 @opindex mas100-syntax
16308 @opindex mno-as100-syntax
16309 When generating assembler output use a syntax that is compatible with
16310 Renesas's AS100 assembler. This syntax can also be handled by the GAS
16311 assembler but it has some restrictions so generating it is not the
16314 @item -mmax-constant-size=@var{N}
16315 @opindex mmax-constant-size
16316 Specifies the maximum size, in bytes, of a constant that can be used as
16317 an operand in a RX instruction. Although the RX instruction set does
16318 allow constants of up to 4 bytes in length to be used in instructions,
16319 a longer value equates to a longer instruction. Thus in some
16320 circumstances it can be beneficial to restrict the size of constants
16321 that are used in instructions. Constants that are too big are instead
16322 placed into a constant pool and referenced via register indirection.
16324 The value @var{N} can be between 0 and 4. A value of 0 (the default)
16325 or 4 means that constants of any size are allowed.
16329 Enable linker relaxation. Linker relaxation is a process whereby the
16330 linker will attempt to reduce the size of a program by finding shorter
16331 versions of various instructions. Disabled by default.
16333 @item -mint-register=@var{N}
16334 @opindex mint-register
16335 Specify the number of registers to reserve for fast interrupt handler
16336 functions. The value @var{N} can be between 0 and 4. A value of 1
16337 means that register @code{r13} will be reserved for the exclusive use
16338 of fast interrupt handlers. A value of 2 reserves @code{r13} and
16339 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
16340 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
16341 A value of 0, the default, does not reserve any registers.
16343 @item -msave-acc-in-interrupts
16344 @opindex msave-acc-in-interrupts
16345 Specifies that interrupt handler functions should preserve the
16346 accumulator register. This is only necessary if normal code might use
16347 the accumulator register, for example because it performs 64-bit
16348 multiplications. The default is to ignore the accumulator as this
16349 makes the interrupt handlers faster.
16353 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
16354 has special significance to the RX port when used with the
16355 @code{interrupt} function attribute. This attribute indicates a
16356 function intended to process fast interrupts. GCC will will ensure
16357 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
16358 and/or @code{r13} and only provided that the normal use of the
16359 corresponding registers have been restricted via the
16360 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
16363 @node S/390 and zSeries Options
16364 @subsection S/390 and zSeries Options
16365 @cindex S/390 and zSeries Options
16367 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
16371 @itemx -msoft-float
16372 @opindex mhard-float
16373 @opindex msoft-float
16374 Use (do not use) the hardware floating-point instructions and registers
16375 for floating-point operations. When @option{-msoft-float} is specified,
16376 functions in @file{libgcc.a} will be used to perform floating-point
16377 operations. When @option{-mhard-float} is specified, the compiler
16378 generates IEEE floating-point instructions. This is the default.
16381 @itemx -mno-hard-dfp
16383 @opindex mno-hard-dfp
16384 Use (do not use) the hardware decimal-floating-point instructions for
16385 decimal-floating-point operations. When @option{-mno-hard-dfp} is
16386 specified, functions in @file{libgcc.a} will be used to perform
16387 decimal-floating-point operations. When @option{-mhard-dfp} is
16388 specified, the compiler generates decimal-floating-point hardware
16389 instructions. This is the default for @option{-march=z9-ec} or higher.
16391 @item -mlong-double-64
16392 @itemx -mlong-double-128
16393 @opindex mlong-double-64
16394 @opindex mlong-double-128
16395 These switches control the size of @code{long double} type. A size
16396 of 64bit makes the @code{long double} type equivalent to the @code{double}
16397 type. This is the default.
16400 @itemx -mno-backchain
16401 @opindex mbackchain
16402 @opindex mno-backchain
16403 Store (do not store) the address of the caller's frame as backchain pointer
16404 into the callee's stack frame.
16405 A backchain may be needed to allow debugging using tools that do not understand
16406 DWARF-2 call frame information.
16407 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
16408 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
16409 the backchain is placed into the topmost word of the 96/160 byte register
16412 In general, code compiled with @option{-mbackchain} is call-compatible with
16413 code compiled with @option{-mmo-backchain}; however, use of the backchain
16414 for debugging purposes usually requires that the whole binary is built with
16415 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
16416 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16417 to build a linux kernel use @option{-msoft-float}.
16419 The default is to not maintain the backchain.
16421 @item -mpacked-stack
16422 @itemx -mno-packed-stack
16423 @opindex mpacked-stack
16424 @opindex mno-packed-stack
16425 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
16426 specified, the compiler uses the all fields of the 96/160 byte register save
16427 area only for their default purpose; unused fields still take up stack space.
16428 When @option{-mpacked-stack} is specified, register save slots are densely
16429 packed at the top of the register save area; unused space is reused for other
16430 purposes, allowing for more efficient use of the available stack space.
16431 However, when @option{-mbackchain} is also in effect, the topmost word of
16432 the save area is always used to store the backchain, and the return address
16433 register is always saved two words below the backchain.
16435 As long as the stack frame backchain is not used, code generated with
16436 @option{-mpacked-stack} is call-compatible with code generated with
16437 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
16438 S/390 or zSeries generated code that uses the stack frame backchain at run
16439 time, not just for debugging purposes. Such code is not call-compatible
16440 with code compiled with @option{-mpacked-stack}. Also, note that the
16441 combination of @option{-mbackchain},
16442 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16443 to build a linux kernel use @option{-msoft-float}.
16445 The default is to not use the packed stack layout.
16448 @itemx -mno-small-exec
16449 @opindex msmall-exec
16450 @opindex mno-small-exec
16451 Generate (or do not generate) code using the @code{bras} instruction
16452 to do subroutine calls.
16453 This only works reliably if the total executable size does not
16454 exceed 64k. The default is to use the @code{basr} instruction instead,
16455 which does not have this limitation.
16461 When @option{-m31} is specified, generate code compliant to the
16462 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
16463 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
16464 particular to generate 64-bit instructions. For the @samp{s390}
16465 targets, the default is @option{-m31}, while the @samp{s390x}
16466 targets default to @option{-m64}.
16472 When @option{-mzarch} is specified, generate code using the
16473 instructions available on z/Architecture.
16474 When @option{-mesa} is specified, generate code using the
16475 instructions available on ESA/390. Note that @option{-mesa} is
16476 not possible with @option{-m64}.
16477 When generating code compliant to the GNU/Linux for S/390 ABI,
16478 the default is @option{-mesa}. When generating code compliant
16479 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
16485 Generate (or do not generate) code using the @code{mvcle} instruction
16486 to perform block moves. When @option{-mno-mvcle} is specified,
16487 use a @code{mvc} loop instead. This is the default unless optimizing for
16494 Print (or do not print) additional debug information when compiling.
16495 The default is to not print debug information.
16497 @item -march=@var{cpu-type}
16499 Generate code that will run on @var{cpu-type}, which is the name of a system
16500 representing a certain processor type. Possible values for
16501 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
16502 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
16503 When generating code using the instructions available on z/Architecture,
16504 the default is @option{-march=z900}. Otherwise, the default is
16505 @option{-march=g5}.
16507 @item -mtune=@var{cpu-type}
16509 Tune to @var{cpu-type} everything applicable about the generated code,
16510 except for the ABI and the set of available instructions.
16511 The list of @var{cpu-type} values is the same as for @option{-march}.
16512 The default is the value used for @option{-march}.
16515 @itemx -mno-tpf-trace
16516 @opindex mtpf-trace
16517 @opindex mno-tpf-trace
16518 Generate code that adds (does not add) in TPF OS specific branches to trace
16519 routines in the operating system. This option is off by default, even
16520 when compiling for the TPF OS@.
16523 @itemx -mno-fused-madd
16524 @opindex mfused-madd
16525 @opindex mno-fused-madd
16526 Generate code that uses (does not use) the floating point multiply and
16527 accumulate instructions. These instructions are generated by default if
16528 hardware floating point is used.
16530 @item -mwarn-framesize=@var{framesize}
16531 @opindex mwarn-framesize
16532 Emit a warning if the current function exceeds the given frame size. Because
16533 this is a compile time check it doesn't need to be a real problem when the program
16534 runs. It is intended to identify functions which most probably cause
16535 a stack overflow. It is useful to be used in an environment with limited stack
16536 size e.g.@: the linux kernel.
16538 @item -mwarn-dynamicstack
16539 @opindex mwarn-dynamicstack
16540 Emit a warning if the function calls alloca or uses dynamically
16541 sized arrays. This is generally a bad idea with a limited stack size.
16543 @item -mstack-guard=@var{stack-guard}
16544 @itemx -mstack-size=@var{stack-size}
16545 @opindex mstack-guard
16546 @opindex mstack-size
16547 If these options are provided the s390 back end emits additional instructions in
16548 the function prologue which trigger a trap if the stack size is @var{stack-guard}
16549 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
16550 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
16551 the frame size of the compiled function is chosen.
16552 These options are intended to be used to help debugging stack overflow problems.
16553 The additionally emitted code causes only little overhead and hence can also be
16554 used in production like systems without greater performance degradation. The given
16555 values have to be exact powers of 2 and @var{stack-size} has to be greater than
16556 @var{stack-guard} without exceeding 64k.
16557 In order to be efficient the extra code makes the assumption that the stack starts
16558 at an address aligned to the value given by @var{stack-size}.
16559 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
16562 @node Score Options
16563 @subsection Score Options
16564 @cindex Score Options
16566 These options are defined for Score implementations:
16571 Compile code for big endian mode. This is the default.
16575 Compile code for little endian mode.
16579 Disable generate bcnz instruction.
16583 Enable generate unaligned load and store instruction.
16587 Enable the use of multiply-accumulate instructions. Disabled by default.
16591 Specify the SCORE5 as the target architecture.
16595 Specify the SCORE5U of the target architecture.
16599 Specify the SCORE7 as the target architecture. This is the default.
16603 Specify the SCORE7D as the target architecture.
16607 @subsection SH Options
16609 These @samp{-m} options are defined for the SH implementations:
16614 Generate code for the SH1.
16618 Generate code for the SH2.
16621 Generate code for the SH2e.
16625 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
16626 that the floating-point unit is not used.
16628 @item -m2a-single-only
16629 @opindex m2a-single-only
16630 Generate code for the SH2a-FPU, in such a way that no double-precision
16631 floating point operations are used.
16634 @opindex m2a-single
16635 Generate code for the SH2a-FPU assuming the floating-point unit is in
16636 single-precision mode by default.
16640 Generate code for the SH2a-FPU assuming the floating-point unit is in
16641 double-precision mode by default.
16645 Generate code for the SH3.
16649 Generate code for the SH3e.
16653 Generate code for the SH4 without a floating-point unit.
16655 @item -m4-single-only
16656 @opindex m4-single-only
16657 Generate code for the SH4 with a floating-point unit that only
16658 supports single-precision arithmetic.
16662 Generate code for the SH4 assuming the floating-point unit is in
16663 single-precision mode by default.
16667 Generate code for the SH4.
16671 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
16672 floating-point unit is not used.
16674 @item -m4a-single-only
16675 @opindex m4a-single-only
16676 Generate code for the SH4a, in such a way that no double-precision
16677 floating point operations are used.
16680 @opindex m4a-single
16681 Generate code for the SH4a assuming the floating-point unit is in
16682 single-precision mode by default.
16686 Generate code for the SH4a.
16690 Same as @option{-m4a-nofpu}, except that it implicitly passes
16691 @option{-dsp} to the assembler. GCC doesn't generate any DSP
16692 instructions at the moment.
16696 Compile code for the processor in big endian mode.
16700 Compile code for the processor in little endian mode.
16704 Align doubles at 64-bit boundaries. Note that this changes the calling
16705 conventions, and thus some functions from the standard C library will
16706 not work unless you recompile it first with @option{-mdalign}.
16710 Shorten some address references at link time, when possible; uses the
16711 linker option @option{-relax}.
16715 Use 32-bit offsets in @code{switch} tables. The default is to use
16720 Enable the use of bit manipulation instructions on SH2A.
16724 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
16725 alignment constraints.
16729 Comply with the calling conventions defined by Renesas.
16733 Comply with the calling conventions defined by Renesas.
16737 Comply with the calling conventions defined for GCC before the Renesas
16738 conventions were available. This option is the default for all
16739 targets of the SH toolchain except for @samp{sh-symbianelf}.
16742 @opindex mnomacsave
16743 Mark the @code{MAC} register as call-clobbered, even if
16744 @option{-mhitachi} is given.
16748 Increase IEEE-compliance of floating-point code.
16749 At the moment, this is equivalent to @option{-fno-finite-math-only}.
16750 When generating 16 bit SH opcodes, getting IEEE-conforming results for
16751 comparisons of NANs / infinities incurs extra overhead in every
16752 floating point comparison, therefore the default is set to
16753 @option{-ffinite-math-only}.
16755 @item -minline-ic_invalidate
16756 @opindex minline-ic_invalidate
16757 Inline code to invalidate instruction cache entries after setting up
16758 nested function trampolines.
16759 This option has no effect if -musermode is in effect and the selected
16760 code generation option (e.g. -m4) does not allow the use of the icbi
16762 If the selected code generation option does not allow the use of the icbi
16763 instruction, and -musermode is not in effect, the inlined code will
16764 manipulate the instruction cache address array directly with an associative
16765 write. This not only requires privileged mode, but it will also
16766 fail if the cache line had been mapped via the TLB and has become unmapped.
16770 Dump instruction size and location in the assembly code.
16773 @opindex mpadstruct
16774 This option is deprecated. It pads structures to multiple of 4 bytes,
16775 which is incompatible with the SH ABI@.
16779 Optimize for space instead of speed. Implied by @option{-Os}.
16782 @opindex mprefergot
16783 When generating position-independent code, emit function calls using
16784 the Global Offset Table instead of the Procedure Linkage Table.
16788 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
16789 if the inlined code would not work in user mode.
16790 This is the default when the target is @code{sh-*-linux*}.
16792 @item -multcost=@var{number}
16793 @opindex multcost=@var{number}
16794 Set the cost to assume for a multiply insn.
16796 @item -mdiv=@var{strategy}
16797 @opindex mdiv=@var{strategy}
16798 Set the division strategy to use for SHmedia code. @var{strategy} must be
16799 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
16800 inv:call2, inv:fp .
16801 "fp" performs the operation in floating point. This has a very high latency,
16802 but needs only a few instructions, so it might be a good choice if
16803 your code has enough easily exploitable ILP to allow the compiler to
16804 schedule the floating point instructions together with other instructions.
16805 Division by zero causes a floating point exception.
16806 "inv" uses integer operations to calculate the inverse of the divisor,
16807 and then multiplies the dividend with the inverse. This strategy allows
16808 cse and hoisting of the inverse calculation. Division by zero calculates
16809 an unspecified result, but does not trap.
16810 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
16811 have been found, or if the entire operation has been hoisted to the same
16812 place, the last stages of the inverse calculation are intertwined with the
16813 final multiply to reduce the overall latency, at the expense of using a few
16814 more instructions, and thus offering fewer scheduling opportunities with
16816 "call" calls a library function that usually implements the inv:minlat
16818 This gives high code density for m5-*media-nofpu compilations.
16819 "call2" uses a different entry point of the same library function, where it
16820 assumes that a pointer to a lookup table has already been set up, which
16821 exposes the pointer load to cse / code hoisting optimizations.
16822 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
16823 code generation, but if the code stays unoptimized, revert to the "call",
16824 "call2", or "fp" strategies, respectively. Note that the
16825 potentially-trapping side effect of division by zero is carried by a
16826 separate instruction, so it is possible that all the integer instructions
16827 are hoisted out, but the marker for the side effect stays where it is.
16828 A recombination to fp operations or a call is not possible in that case.
16829 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
16830 that the inverse calculation was nor separated from the multiply, they speed
16831 up division where the dividend fits into 20 bits (plus sign where applicable),
16832 by inserting a test to skip a number of operations in this case; this test
16833 slows down the case of larger dividends. inv20u assumes the case of a such
16834 a small dividend to be unlikely, and inv20l assumes it to be likely.
16836 @item -maccumulate-outgoing-args
16837 @opindex maccumulate-outgoing-args
16838 Reserve space once for outgoing arguments in the function prologue rather
16839 than around each call. Generally beneficial for performance and size. Also
16840 needed for unwinding to avoid changing the stack frame around conditional code.
16842 @item -mdivsi3_libfunc=@var{name}
16843 @opindex mdivsi3_libfunc=@var{name}
16844 Set the name of the library function used for 32 bit signed division to
16845 @var{name}. This only affect the name used in the call and inv:call
16846 division strategies, and the compiler will still expect the same
16847 sets of input/output/clobbered registers as if this option was not present.
16849 @item -mfixed-range=@var{register-range}
16850 @opindex mfixed-range
16851 Generate code treating the given register range as fixed registers.
16852 A fixed register is one that the register allocator can not use. This is
16853 useful when compiling kernel code. A register range is specified as
16854 two registers separated by a dash. Multiple register ranges can be
16855 specified separated by a comma.
16857 @item -madjust-unroll
16858 @opindex madjust-unroll
16859 Throttle unrolling to avoid thrashing target registers.
16860 This option only has an effect if the gcc code base supports the
16861 TARGET_ADJUST_UNROLL_MAX target hook.
16863 @item -mindexed-addressing
16864 @opindex mindexed-addressing
16865 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
16866 This is only safe if the hardware and/or OS implement 32 bit wrap-around
16867 semantics for the indexed addressing mode. The architecture allows the
16868 implementation of processors with 64 bit MMU, which the OS could use to
16869 get 32 bit addressing, but since no current hardware implementation supports
16870 this or any other way to make the indexed addressing mode safe to use in
16871 the 32 bit ABI, the default is -mno-indexed-addressing.
16873 @item -mgettrcost=@var{number}
16874 @opindex mgettrcost=@var{number}
16875 Set the cost assumed for the gettr instruction to @var{number}.
16876 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
16880 Assume pt* instructions won't trap. This will generally generate better
16881 scheduled code, but is unsafe on current hardware. The current architecture
16882 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
16883 This has the unintentional effect of making it unsafe to schedule ptabs /
16884 ptrel before a branch, or hoist it out of a loop. For example,
16885 __do_global_ctors, a part of libgcc that runs constructors at program
16886 startup, calls functions in a list which is delimited by @minus{}1. With the
16887 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
16888 That means that all the constructors will be run a bit quicker, but when
16889 the loop comes to the end of the list, the program crashes because ptabs
16890 loads @minus{}1 into a target register. Since this option is unsafe for any
16891 hardware implementing the current architecture specification, the default
16892 is -mno-pt-fixed. Unless the user specifies a specific cost with
16893 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
16894 this deters register allocation using target registers for storing
16897 @item -minvalid-symbols
16898 @opindex minvalid-symbols
16899 Assume symbols might be invalid. Ordinary function symbols generated by
16900 the compiler will always be valid to load with movi/shori/ptabs or
16901 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
16902 to generate symbols that will cause ptabs / ptrel to trap.
16903 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
16904 It will then prevent cross-basic-block cse, hoisting and most scheduling
16905 of symbol loads. The default is @option{-mno-invalid-symbols}.
16908 @node Solaris 2 Options
16909 @subsection Solaris 2 Options
16910 @cindex Solaris 2 options
16912 These @samp{-m} options are supported on Solaris 2:
16915 @item -mimpure-text
16916 @opindex mimpure-text
16917 @option{-mimpure-text}, used in addition to @option{-shared}, tells
16918 the compiler to not pass @option{-z text} to the linker when linking a
16919 shared object. Using this option, you can link position-dependent
16920 code into a shared object.
16922 @option{-mimpure-text} suppresses the ``relocations remain against
16923 allocatable but non-writable sections'' linker error message.
16924 However, the necessary relocations will trigger copy-on-write, and the
16925 shared object is not actually shared across processes. Instead of
16926 using @option{-mimpure-text}, you should compile all source code with
16927 @option{-fpic} or @option{-fPIC}.
16931 These switches are supported in addition to the above on Solaris 2:
16936 Add support for multithreading using the Solaris threads library. This
16937 option sets flags for both the preprocessor and linker. This option does
16938 not affect the thread safety of object code produced by the compiler or
16939 that of libraries supplied with it.
16943 Add support for multithreading using the POSIX threads library. This
16944 option sets flags for both the preprocessor and linker. This option does
16945 not affect the thread safety of object code produced by the compiler or
16946 that of libraries supplied with it.
16950 This is a synonym for @option{-pthreads}.
16953 @node SPARC Options
16954 @subsection SPARC Options
16955 @cindex SPARC options
16957 These @samp{-m} options are supported on the SPARC:
16960 @item -mno-app-regs
16962 @opindex mno-app-regs
16964 Specify @option{-mapp-regs} to generate output using the global registers
16965 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
16968 To be fully SVR4 ABI compliant at the cost of some performance loss,
16969 specify @option{-mno-app-regs}. You should compile libraries and system
16970 software with this option.
16973 @itemx -mhard-float
16975 @opindex mhard-float
16976 Generate output containing floating point instructions. This is the
16980 @itemx -msoft-float
16982 @opindex msoft-float
16983 Generate output containing library calls for floating point.
16984 @strong{Warning:} the requisite libraries are not available for all SPARC
16985 targets. Normally the facilities of the machine's usual C compiler are
16986 used, but this cannot be done directly in cross-compilation. You must make
16987 your own arrangements to provide suitable library functions for
16988 cross-compilation. The embedded targets @samp{sparc-*-aout} and
16989 @samp{sparclite-*-*} do provide software floating point support.
16991 @option{-msoft-float} changes the calling convention in the output file;
16992 therefore, it is only useful if you compile @emph{all} of a program with
16993 this option. In particular, you need to compile @file{libgcc.a}, the
16994 library that comes with GCC, with @option{-msoft-float} in order for
16997 @item -mhard-quad-float
16998 @opindex mhard-quad-float
16999 Generate output containing quad-word (long double) floating point
17002 @item -msoft-quad-float
17003 @opindex msoft-quad-float
17004 Generate output containing library calls for quad-word (long double)
17005 floating point instructions. The functions called are those specified
17006 in the SPARC ABI@. This is the default.
17008 As of this writing, there are no SPARC implementations that have hardware
17009 support for the quad-word floating point instructions. They all invoke
17010 a trap handler for one of these instructions, and then the trap handler
17011 emulates the effect of the instruction. Because of the trap handler overhead,
17012 this is much slower than calling the ABI library routines. Thus the
17013 @option{-msoft-quad-float} option is the default.
17015 @item -mno-unaligned-doubles
17016 @itemx -munaligned-doubles
17017 @opindex mno-unaligned-doubles
17018 @opindex munaligned-doubles
17019 Assume that doubles have 8 byte alignment. This is the default.
17021 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
17022 alignment only if they are contained in another type, or if they have an
17023 absolute address. Otherwise, it assumes they have 4 byte alignment.
17024 Specifying this option avoids some rare compatibility problems with code
17025 generated by other compilers. It is not the default because it results
17026 in a performance loss, especially for floating point code.
17028 @item -mno-faster-structs
17029 @itemx -mfaster-structs
17030 @opindex mno-faster-structs
17031 @opindex mfaster-structs
17032 With @option{-mfaster-structs}, the compiler assumes that structures
17033 should have 8 byte alignment. This enables the use of pairs of
17034 @code{ldd} and @code{std} instructions for copies in structure
17035 assignment, in place of twice as many @code{ld} and @code{st} pairs.
17036 However, the use of this changed alignment directly violates the SPARC
17037 ABI@. Thus, it's intended only for use on targets where the developer
17038 acknowledges that their resulting code will not be directly in line with
17039 the rules of the ABI@.
17041 @item -mcpu=@var{cpu_type}
17043 Set the instruction set, register set, and instruction scheduling parameters
17044 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
17045 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
17046 @samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
17047 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
17048 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
17050 Default instruction scheduling parameters are used for values that select
17051 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
17052 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
17054 Here is a list of each supported architecture and their supported
17059 v8: supersparc, hypersparc, leon
17060 sparclite: f930, f934, sparclite86x
17062 v9: ultrasparc, ultrasparc3, niagara, niagara2
17065 By default (unless configured otherwise), GCC generates code for the V7
17066 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
17067 additionally optimizes it for the Cypress CY7C602 chip, as used in the
17068 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
17069 SPARCStation 1, 2, IPX etc.
17071 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
17072 architecture. The only difference from V7 code is that the compiler emits
17073 the integer multiply and integer divide instructions which exist in SPARC-V8
17074 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
17075 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
17078 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
17079 the SPARC architecture. This adds the integer multiply, integer divide step
17080 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
17081 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
17082 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
17083 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
17084 MB86934 chip, which is the more recent SPARClite with FPU@.
17086 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
17087 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
17088 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
17089 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
17090 optimizes it for the TEMIC SPARClet chip.
17092 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
17093 architecture. This adds 64-bit integer and floating-point move instructions,
17094 3 additional floating-point condition code registers and conditional move
17095 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
17096 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
17097 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
17098 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
17099 @option{-mcpu=niagara}, the compiler additionally optimizes it for
17100 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
17101 additionally optimizes it for Sun UltraSPARC T2 chips.
17103 @item -mtune=@var{cpu_type}
17105 Set the instruction scheduling parameters for machine type
17106 @var{cpu_type}, but do not set the instruction set or register set that the
17107 option @option{-mcpu=@var{cpu_type}} would.
17109 The same values for @option{-mcpu=@var{cpu_type}} can be used for
17110 @option{-mtune=@var{cpu_type}}, but the only useful values are those
17111 that select a particular CPU implementation. Those are @samp{cypress},
17112 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934},
17113 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3},
17114 @samp{niagara}, and @samp{niagara2}.
17119 @opindex mno-v8plus
17120 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
17121 difference from the V8 ABI is that the global and out registers are
17122 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
17123 mode for all SPARC-V9 processors.
17129 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
17130 Visual Instruction Set extensions. The default is @option{-mno-vis}.
17133 These @samp{-m} options are supported in addition to the above
17134 on SPARC-V9 processors in 64-bit environments:
17137 @item -mlittle-endian
17138 @opindex mlittle-endian
17139 Generate code for a processor running in little-endian mode. It is only
17140 available for a few configurations and most notably not on Solaris and Linux.
17146 Generate code for a 32-bit or 64-bit environment.
17147 The 32-bit environment sets int, long and pointer to 32 bits.
17148 The 64-bit environment sets int to 32 bits and long and pointer
17151 @item -mcmodel=medlow
17152 @opindex mcmodel=medlow
17153 Generate code for the Medium/Low code model: 64-bit addresses, programs
17154 must be linked in the low 32 bits of memory. Programs can be statically
17155 or dynamically linked.
17157 @item -mcmodel=medmid
17158 @opindex mcmodel=medmid
17159 Generate code for the Medium/Middle code model: 64-bit addresses, programs
17160 must be linked in the low 44 bits of memory, the text and data segments must
17161 be less than 2GB in size and the data segment must be located within 2GB of
17164 @item -mcmodel=medany
17165 @opindex mcmodel=medany
17166 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
17167 may be linked anywhere in memory, the text and data segments must be less
17168 than 2GB in size and the data segment must be located within 2GB of the
17171 @item -mcmodel=embmedany
17172 @opindex mcmodel=embmedany
17173 Generate code for the Medium/Anywhere code model for embedded systems:
17174 64-bit addresses, the text and data segments must be less than 2GB in
17175 size, both starting anywhere in memory (determined at link time). The
17176 global register %g4 points to the base of the data segment. Programs
17177 are statically linked and PIC is not supported.
17180 @itemx -mno-stack-bias
17181 @opindex mstack-bias
17182 @opindex mno-stack-bias
17183 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
17184 frame pointer if present, are offset by @minus{}2047 which must be added back
17185 when making stack frame references. This is the default in 64-bit mode.
17186 Otherwise, assume no such offset is present.
17190 @subsection SPU Options
17191 @cindex SPU options
17193 These @samp{-m} options are supported on the SPU:
17197 @itemx -merror-reloc
17198 @opindex mwarn-reloc
17199 @opindex merror-reloc
17201 The loader for SPU does not handle dynamic relocations. By default, GCC
17202 will give an error when it generates code that requires a dynamic
17203 relocation. @option{-mno-error-reloc} disables the error,
17204 @option{-mwarn-reloc} will generate a warning instead.
17207 @itemx -munsafe-dma
17209 @opindex munsafe-dma
17211 Instructions which initiate or test completion of DMA must not be
17212 reordered with respect to loads and stores of the memory which is being
17213 accessed. Users typically address this problem using the volatile
17214 keyword, but that can lead to inefficient code in places where the
17215 memory is known to not change. Rather than mark the memory as volatile
17216 we treat the DMA instructions as potentially effecting all memory. With
17217 @option{-munsafe-dma} users must use the volatile keyword to protect
17220 @item -mbranch-hints
17221 @opindex mbranch-hints
17223 By default, GCC will generate a branch hint instruction to avoid
17224 pipeline stalls for always taken or probably taken branches. A hint
17225 will not be generated closer than 8 instructions away from its branch.
17226 There is little reason to disable them, except for debugging purposes,
17227 or to make an object a little bit smaller.
17231 @opindex msmall-mem
17232 @opindex mlarge-mem
17234 By default, GCC generates code assuming that addresses are never larger
17235 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
17236 a full 32 bit address.
17241 By default, GCC links against startup code that assumes the SPU-style
17242 main function interface (which has an unconventional parameter list).
17243 With @option{-mstdmain}, GCC will link your program against startup
17244 code that assumes a C99-style interface to @code{main}, including a
17245 local copy of @code{argv} strings.
17247 @item -mfixed-range=@var{register-range}
17248 @opindex mfixed-range
17249 Generate code treating the given register range as fixed registers.
17250 A fixed register is one that the register allocator can not use. This is
17251 useful when compiling kernel code. A register range is specified as
17252 two registers separated by a dash. Multiple register ranges can be
17253 specified separated by a comma.
17259 Compile code assuming that pointers to the PPU address space accessed
17260 via the @code{__ea} named address space qualifier are either 32 or 64
17261 bits wide. The default is 32 bits. As this is an ABI changing option,
17262 all object code in an executable must be compiled with the same setting.
17264 @item -maddress-space-conversion
17265 @itemx -mno-address-space-conversion
17266 @opindex maddress-space-conversion
17267 @opindex mno-address-space-conversion
17268 Allow/disallow treating the @code{__ea} address space as superset
17269 of the generic address space. This enables explicit type casts
17270 between @code{__ea} and generic pointer as well as implicit
17271 conversions of generic pointers to @code{__ea} pointers. The
17272 default is to allow address space pointer conversions.
17274 @item -mcache-size=@var{cache-size}
17275 @opindex mcache-size
17276 This option controls the version of libgcc that the compiler links to an
17277 executable and selects a software-managed cache for accessing variables
17278 in the @code{__ea} address space with a particular cache size. Possible
17279 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
17280 and @samp{128}. The default cache size is 64KB.
17282 @item -matomic-updates
17283 @itemx -mno-atomic-updates
17284 @opindex matomic-updates
17285 @opindex mno-atomic-updates
17286 This option controls the version of libgcc that the compiler links to an
17287 executable and selects whether atomic updates to the software-managed
17288 cache of PPU-side variables are used. If you use atomic updates, changes
17289 to a PPU variable from SPU code using the @code{__ea} named address space
17290 qualifier will not interfere with changes to other PPU variables residing
17291 in the same cache line from PPU code. If you do not use atomic updates,
17292 such interference may occur; however, writing back cache lines will be
17293 more efficient. The default behavior is to use atomic updates.
17296 @itemx -mdual-nops=@var{n}
17297 @opindex mdual-nops
17298 By default, GCC will insert nops to increase dual issue when it expects
17299 it to increase performance. @var{n} can be a value from 0 to 10. A
17300 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
17301 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
17303 @item -mhint-max-nops=@var{n}
17304 @opindex mhint-max-nops
17305 Maximum number of nops to insert for a branch hint. A branch hint must
17306 be at least 8 instructions away from the branch it is effecting. GCC
17307 will insert up to @var{n} nops to enforce this, otherwise it will not
17308 generate the branch hint.
17310 @item -mhint-max-distance=@var{n}
17311 @opindex mhint-max-distance
17312 The encoding of the branch hint instruction limits the hint to be within
17313 256 instructions of the branch it is effecting. By default, GCC makes
17314 sure it is within 125.
17317 @opindex msafe-hints
17318 Work around a hardware bug which causes the SPU to stall indefinitely.
17319 By default, GCC will insert the @code{hbrp} instruction to make sure
17320 this stall won't happen.
17324 @node System V Options
17325 @subsection Options for System V
17327 These additional options are available on System V Release 4 for
17328 compatibility with other compilers on those systems:
17333 Create a shared object.
17334 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
17338 Identify the versions of each tool used by the compiler, in a
17339 @code{.ident} assembler directive in the output.
17343 Refrain from adding @code{.ident} directives to the output file (this is
17346 @item -YP,@var{dirs}
17348 Search the directories @var{dirs}, and no others, for libraries
17349 specified with @option{-l}.
17351 @item -Ym,@var{dir}
17353 Look in the directory @var{dir} to find the M4 preprocessor.
17354 The assembler uses this option.
17355 @c This is supposed to go with a -Yd for predefined M4 macro files, but
17356 @c the generic assembler that comes with Solaris takes just -Ym.
17360 @subsection V850 Options
17361 @cindex V850 Options
17363 These @samp{-m} options are defined for V850 implementations:
17367 @itemx -mno-long-calls
17368 @opindex mlong-calls
17369 @opindex mno-long-calls
17370 Treat all calls as being far away (near). If calls are assumed to be
17371 far away, the compiler will always load the functions address up into a
17372 register, and call indirect through the pointer.
17378 Do not optimize (do optimize) basic blocks that use the same index
17379 pointer 4 or more times to copy pointer into the @code{ep} register, and
17380 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
17381 option is on by default if you optimize.
17383 @item -mno-prolog-function
17384 @itemx -mprolog-function
17385 @opindex mno-prolog-function
17386 @opindex mprolog-function
17387 Do not use (do use) external functions to save and restore registers
17388 at the prologue and epilogue of a function. The external functions
17389 are slower, but use less code space if more than one function saves
17390 the same number of registers. The @option{-mprolog-function} option
17391 is on by default if you optimize.
17395 Try to make the code as small as possible. At present, this just turns
17396 on the @option{-mep} and @option{-mprolog-function} options.
17398 @item -mtda=@var{n}
17400 Put static or global variables whose size is @var{n} bytes or less into
17401 the tiny data area that register @code{ep} points to. The tiny data
17402 area can hold up to 256 bytes in total (128 bytes for byte references).
17404 @item -msda=@var{n}
17406 Put static or global variables whose size is @var{n} bytes or less into
17407 the small data area that register @code{gp} points to. The small data
17408 area can hold up to 64 kilobytes.
17410 @item -mzda=@var{n}
17412 Put static or global variables whose size is @var{n} bytes or less into
17413 the first 32 kilobytes of memory.
17417 Specify that the target processor is the V850.
17420 @opindex mbig-switch
17421 Generate code suitable for big switch tables. Use this option only if
17422 the assembler/linker complain about out of range branches within a switch
17427 This option will cause r2 and r5 to be used in the code generated by
17428 the compiler. This setting is the default.
17430 @item -mno-app-regs
17431 @opindex mno-app-regs
17432 This option will cause r2 and r5 to be treated as fixed registers.
17436 Specify that the target processor is the V850E2V3. The preprocessor
17437 constants @samp{__v850e2v3__} will be defined if
17438 this option is used.
17442 Specify that the target processor is the V850E2. The preprocessor
17443 constants @samp{__v850e2__} will be defined if
17447 Specify that the target processor is the V850E1. The preprocessor
17448 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
17452 Specify that the target processor is the V850ES. This is an alias for
17453 the @option{-mv850e1} option.
17457 Specify that the target processor is the V850E@. The preprocessor
17458 constant @samp{__v850e__} will be defined if this option is used.
17460 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
17461 nor @option{-mv850e2} nor @option{-mv850e2v3}
17462 are defined then a default target processor will be chosen and the
17463 relevant @samp{__v850*__} preprocessor constant will be defined.
17465 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
17466 defined, regardless of which processor variant is the target.
17468 @item -mdisable-callt
17469 @opindex mdisable-callt
17470 This option will suppress generation of the CALLT instruction for the
17471 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
17472 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
17477 @subsection VAX Options
17478 @cindex VAX options
17480 These @samp{-m} options are defined for the VAX:
17485 Do not output certain jump instructions (@code{aobleq} and so on)
17486 that the Unix assembler for the VAX cannot handle across long
17491 Do output those jump instructions, on the assumption that you
17492 will assemble with the GNU assembler.
17496 Output code for g-format floating point numbers instead of d-format.
17499 @node VxWorks Options
17500 @subsection VxWorks Options
17501 @cindex VxWorks Options
17503 The options in this section are defined for all VxWorks targets.
17504 Options specific to the target hardware are listed with the other
17505 options for that target.
17510 GCC can generate code for both VxWorks kernels and real time processes
17511 (RTPs). This option switches from the former to the latter. It also
17512 defines the preprocessor macro @code{__RTP__}.
17515 @opindex non-static
17516 Link an RTP executable against shared libraries rather than static
17517 libraries. The options @option{-static} and @option{-shared} can
17518 also be used for RTPs (@pxref{Link Options}); @option{-static}
17525 These options are passed down to the linker. They are defined for
17526 compatibility with Diab.
17529 @opindex Xbind-lazy
17530 Enable lazy binding of function calls. This option is equivalent to
17531 @option{-Wl,-z,now} and is defined for compatibility with Diab.
17535 Disable lazy binding of function calls. This option is the default and
17536 is defined for compatibility with Diab.
17539 @node x86-64 Options
17540 @subsection x86-64 Options
17541 @cindex x86-64 options
17543 These are listed under @xref{i386 and x86-64 Options}.
17545 @node i386 and x86-64 Windows Options
17546 @subsection i386 and x86-64 Windows Options
17547 @cindex i386 and x86-64 Windows Options
17549 These additional options are available for Windows targets:
17554 This option is available for Cygwin and MinGW targets. It
17555 specifies that a console application is to be generated, by
17556 instructing the linker to set the PE header subsystem type
17557 required for console applications.
17558 This is the default behavior for Cygwin and MinGW targets.
17562 This option is available for Cygwin and MinGW targets. It
17563 specifies that a DLL - a dynamic link library - is to be
17564 generated, enabling the selection of the required runtime
17565 startup object and entry point.
17567 @item -mnop-fun-dllimport
17568 @opindex mnop-fun-dllimport
17569 This option is available for Cygwin and MinGW targets. It
17570 specifies that the dllimport attribute should be ignored.
17574 This option is available for MinGW targets. It specifies
17575 that MinGW-specific thread support is to be used.
17579 This option is available for mingw-w64 targets. It specifies
17580 that the UNICODE macro is getting pre-defined and that the
17581 unicode capable runtime startup code is chosen.
17585 This option is available for Cygwin and MinGW targets. It
17586 specifies that the typical Windows pre-defined macros are to
17587 be set in the pre-processor, but does not influence the choice
17588 of runtime library/startup code.
17592 This option is available for Cygwin and MinGW targets. It
17593 specifies that a GUI application is to be generated by
17594 instructing the linker to set the PE header subsystem type
17597 @item -fno-set-stack-executable
17598 @opindex fno-set-stack-executable
17599 This option is available for MinGW targets. It specifies that
17600 the executable flag for stack used by nested functions isn't
17601 set. This is necessary for binaries running in kernel mode of
17602 Windows, as there the user32 API, which is used to set executable
17603 privileges, isn't available.
17605 @item -mpe-aligned-commons
17606 @opindex mpe-aligned-commons
17607 This option is available for Cygwin and MinGW targets. It
17608 specifies that the GNU extension to the PE file format that
17609 permits the correct alignment of COMMON variables should be
17610 used when generating code. It will be enabled by default if
17611 GCC detects that the target assembler found during configuration
17612 supports the feature.
17615 See also under @ref{i386 and x86-64 Options} for standard options.
17617 @node Xstormy16 Options
17618 @subsection Xstormy16 Options
17619 @cindex Xstormy16 Options
17621 These options are defined for Xstormy16:
17626 Choose startup files and linker script suitable for the simulator.
17629 @node Xtensa Options
17630 @subsection Xtensa Options
17631 @cindex Xtensa Options
17633 These options are supported for Xtensa targets:
17637 @itemx -mno-const16
17639 @opindex mno-const16
17640 Enable or disable use of @code{CONST16} instructions for loading
17641 constant values. The @code{CONST16} instruction is currently not a
17642 standard option from Tensilica. When enabled, @code{CONST16}
17643 instructions are always used in place of the standard @code{L32R}
17644 instructions. The use of @code{CONST16} is enabled by default only if
17645 the @code{L32R} instruction is not available.
17648 @itemx -mno-fused-madd
17649 @opindex mfused-madd
17650 @opindex mno-fused-madd
17651 Enable or disable use of fused multiply/add and multiply/subtract
17652 instructions in the floating-point option. This has no effect if the
17653 floating-point option is not also enabled. Disabling fused multiply/add
17654 and multiply/subtract instructions forces the compiler to use separate
17655 instructions for the multiply and add/subtract operations. This may be
17656 desirable in some cases where strict IEEE 754-compliant results are
17657 required: the fused multiply add/subtract instructions do not round the
17658 intermediate result, thereby producing results with @emph{more} bits of
17659 precision than specified by the IEEE standard. Disabling fused multiply
17660 add/subtract instructions also ensures that the program output is not
17661 sensitive to the compiler's ability to combine multiply and add/subtract
17664 @item -mserialize-volatile
17665 @itemx -mno-serialize-volatile
17666 @opindex mserialize-volatile
17667 @opindex mno-serialize-volatile
17668 When this option is enabled, GCC inserts @code{MEMW} instructions before
17669 @code{volatile} memory references to guarantee sequential consistency.
17670 The default is @option{-mserialize-volatile}. Use
17671 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
17673 @item -mforce-no-pic
17674 @opindex mforce-no-pic
17675 For targets, like GNU/Linux, where all user-mode Xtensa code must be
17676 position-independent code (PIC), this option disables PIC for compiling
17679 @item -mtext-section-literals
17680 @itemx -mno-text-section-literals
17681 @opindex mtext-section-literals
17682 @opindex mno-text-section-literals
17683 Control the treatment of literal pools. The default is
17684 @option{-mno-text-section-literals}, which places literals in a separate
17685 section in the output file. This allows the literal pool to be placed
17686 in a data RAM/ROM, and it also allows the linker to combine literal
17687 pools from separate object files to remove redundant literals and
17688 improve code size. With @option{-mtext-section-literals}, the literals
17689 are interspersed in the text section in order to keep them as close as
17690 possible to their references. This may be necessary for large assembly
17693 @item -mtarget-align
17694 @itemx -mno-target-align
17695 @opindex mtarget-align
17696 @opindex mno-target-align
17697 When this option is enabled, GCC instructs the assembler to
17698 automatically align instructions to reduce branch penalties at the
17699 expense of some code density. The assembler attempts to widen density
17700 instructions to align branch targets and the instructions following call
17701 instructions. If there are not enough preceding safe density
17702 instructions to align a target, no widening will be performed. The
17703 default is @option{-mtarget-align}. These options do not affect the
17704 treatment of auto-aligned instructions like @code{LOOP}, which the
17705 assembler will always align, either by widening density instructions or
17706 by inserting no-op instructions.
17709 @itemx -mno-longcalls
17710 @opindex mlongcalls
17711 @opindex mno-longcalls
17712 When this option is enabled, GCC instructs the assembler to translate
17713 direct calls to indirect calls unless it can determine that the target
17714 of a direct call is in the range allowed by the call instruction. This
17715 translation typically occurs for calls to functions in other source
17716 files. Specifically, the assembler translates a direct @code{CALL}
17717 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
17718 The default is @option{-mno-longcalls}. This option should be used in
17719 programs where the call target can potentially be out of range. This
17720 option is implemented in the assembler, not the compiler, so the
17721 assembly code generated by GCC will still show direct call
17722 instructions---look at the disassembled object code to see the actual
17723 instructions. Note that the assembler will use an indirect call for
17724 every cross-file call, not just those that really will be out of range.
17727 @node zSeries Options
17728 @subsection zSeries Options
17729 @cindex zSeries options
17731 These are listed under @xref{S/390 and zSeries Options}.
17733 @node Code Gen Options
17734 @section Options for Code Generation Conventions
17735 @cindex code generation conventions
17736 @cindex options, code generation
17737 @cindex run-time options
17739 These machine-independent options control the interface conventions
17740 used in code generation.
17742 Most of them have both positive and negative forms; the negative form
17743 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
17744 one of the forms is listed---the one which is not the default. You
17745 can figure out the other form by either removing @samp{no-} or adding
17749 @item -fbounds-check
17750 @opindex fbounds-check
17751 For front-ends that support it, generate additional code to check that
17752 indices used to access arrays are within the declared range. This is
17753 currently only supported by the Java and Fortran front-ends, where
17754 this option defaults to true and false respectively.
17758 This option generates traps for signed overflow on addition, subtraction,
17759 multiplication operations.
17763 This option instructs the compiler to assume that signed arithmetic
17764 overflow of addition, subtraction and multiplication wraps around
17765 using twos-complement representation. This flag enables some optimizations
17766 and disables others. This option is enabled by default for the Java
17767 front-end, as required by the Java language specification.
17770 @opindex fexceptions
17771 Enable exception handling. Generates extra code needed to propagate
17772 exceptions. For some targets, this implies GCC will generate frame
17773 unwind information for all functions, which can produce significant data
17774 size overhead, although it does not affect execution. If you do not
17775 specify this option, GCC will enable it by default for languages like
17776 C++ which normally require exception handling, and disable it for
17777 languages like C that do not normally require it. However, you may need
17778 to enable this option when compiling C code that needs to interoperate
17779 properly with exception handlers written in C++. You may also wish to
17780 disable this option if you are compiling older C++ programs that don't
17781 use exception handling.
17783 @item -fnon-call-exceptions
17784 @opindex fnon-call-exceptions
17785 Generate code that allows trapping instructions to throw exceptions.
17786 Note that this requires platform-specific runtime support that does
17787 not exist everywhere. Moreover, it only allows @emph{trapping}
17788 instructions to throw exceptions, i.e.@: memory references or floating
17789 point instructions. It does not allow exceptions to be thrown from
17790 arbitrary signal handlers such as @code{SIGALRM}.
17792 @item -funwind-tables
17793 @opindex funwind-tables
17794 Similar to @option{-fexceptions}, except that it will just generate any needed
17795 static data, but will not affect the generated code in any other way.
17796 You will normally not enable this option; instead, a language processor
17797 that needs this handling would enable it on your behalf.
17799 @item -fasynchronous-unwind-tables
17800 @opindex fasynchronous-unwind-tables
17801 Generate unwind table in dwarf2 format, if supported by target machine. The
17802 table is exact at each instruction boundary, so it can be used for stack
17803 unwinding from asynchronous events (such as debugger or garbage collector).
17805 @item -fpcc-struct-return
17806 @opindex fpcc-struct-return
17807 Return ``short'' @code{struct} and @code{union} values in memory like
17808 longer ones, rather than in registers. This convention is less
17809 efficient, but it has the advantage of allowing intercallability between
17810 GCC-compiled files and files compiled with other compilers, particularly
17811 the Portable C Compiler (pcc).
17813 The precise convention for returning structures in memory depends
17814 on the target configuration macros.
17816 Short structures and unions are those whose size and alignment match
17817 that of some integer type.
17819 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
17820 switch is not binary compatible with code compiled with the
17821 @option{-freg-struct-return} switch.
17822 Use it to conform to a non-default application binary interface.
17824 @item -freg-struct-return
17825 @opindex freg-struct-return
17826 Return @code{struct} and @code{union} values in registers when possible.
17827 This is more efficient for small structures than
17828 @option{-fpcc-struct-return}.
17830 If you specify neither @option{-fpcc-struct-return} nor
17831 @option{-freg-struct-return}, GCC defaults to whichever convention is
17832 standard for the target. If there is no standard convention, GCC
17833 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
17834 the principal compiler. In those cases, we can choose the standard, and
17835 we chose the more efficient register return alternative.
17837 @strong{Warning:} code compiled with the @option{-freg-struct-return}
17838 switch is not binary compatible with code compiled with the
17839 @option{-fpcc-struct-return} switch.
17840 Use it to conform to a non-default application binary interface.
17842 @item -fshort-enums
17843 @opindex fshort-enums
17844 Allocate to an @code{enum} type only as many bytes as it needs for the
17845 declared range of possible values. Specifically, the @code{enum} type
17846 will be equivalent to the smallest integer type which has enough room.
17848 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
17849 code that is not binary compatible with code generated without that switch.
17850 Use it to conform to a non-default application binary interface.
17852 @item -fshort-double
17853 @opindex fshort-double
17854 Use the same size for @code{double} as for @code{float}.
17856 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
17857 code that is not binary compatible with code generated without that switch.
17858 Use it to conform to a non-default application binary interface.
17860 @item -fshort-wchar
17861 @opindex fshort-wchar
17862 Override the underlying type for @samp{wchar_t} to be @samp{short
17863 unsigned int} instead of the default for the target. This option is
17864 useful for building programs to run under WINE@.
17866 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
17867 code that is not binary compatible with code generated without that switch.
17868 Use it to conform to a non-default application binary interface.
17871 @opindex fno-common
17872 In C code, controls the placement of uninitialized global variables.
17873 Unix C compilers have traditionally permitted multiple definitions of
17874 such variables in different compilation units by placing the variables
17876 This is the behavior specified by @option{-fcommon}, and is the default
17877 for GCC on most targets.
17878 On the other hand, this behavior is not required by ISO C, and on some
17879 targets may carry a speed or code size penalty on variable references.
17880 The @option{-fno-common} option specifies that the compiler should place
17881 uninitialized global variables in the data section of the object file,
17882 rather than generating them as common blocks.
17883 This has the effect that if the same variable is declared
17884 (without @code{extern}) in two different compilations,
17885 you will get a multiple-definition error when you link them.
17886 In this case, you must compile with @option{-fcommon} instead.
17887 Compiling with @option{-fno-common} is useful on targets for which
17888 it provides better performance, or if you wish to verify that the
17889 program will work on other systems which always treat uninitialized
17890 variable declarations this way.
17894 Ignore the @samp{#ident} directive.
17896 @item -finhibit-size-directive
17897 @opindex finhibit-size-directive
17898 Don't output a @code{.size} assembler directive, or anything else that
17899 would cause trouble if the function is split in the middle, and the
17900 two halves are placed at locations far apart in memory. This option is
17901 used when compiling @file{crtstuff.c}; you should not need to use it
17904 @item -fverbose-asm
17905 @opindex fverbose-asm
17906 Put extra commentary information in the generated assembly code to
17907 make it more readable. This option is generally only of use to those
17908 who actually need to read the generated assembly code (perhaps while
17909 debugging the compiler itself).
17911 @option{-fno-verbose-asm}, the default, causes the
17912 extra information to be omitted and is useful when comparing two assembler
17915 @item -frecord-gcc-switches
17916 @opindex frecord-gcc-switches
17917 This switch causes the command line that was used to invoke the
17918 compiler to be recorded into the object file that is being created.
17919 This switch is only implemented on some targets and the exact format
17920 of the recording is target and binary file format dependent, but it
17921 usually takes the form of a section containing ASCII text. This
17922 switch is related to the @option{-fverbose-asm} switch, but that
17923 switch only records information in the assembler output file as
17924 comments, so it never reaches the object file.
17928 @cindex global offset table
17930 Generate position-independent code (PIC) suitable for use in a shared
17931 library, if supported for the target machine. Such code accesses all
17932 constant addresses through a global offset table (GOT)@. The dynamic
17933 loader resolves the GOT entries when the program starts (the dynamic
17934 loader is not part of GCC; it is part of the operating system). If
17935 the GOT size for the linked executable exceeds a machine-specific
17936 maximum size, you get an error message from the linker indicating that
17937 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
17938 instead. (These maximums are 8k on the SPARC and 32k
17939 on the m68k and RS/6000. The 386 has no such limit.)
17941 Position-independent code requires special support, and therefore works
17942 only on certain machines. For the 386, GCC supports PIC for System V
17943 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
17944 position-independent.
17946 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17951 If supported for the target machine, emit position-independent code,
17952 suitable for dynamic linking and avoiding any limit on the size of the
17953 global offset table. This option makes a difference on the m68k,
17954 PowerPC and SPARC@.
17956 Position-independent code requires special support, and therefore works
17957 only on certain machines.
17959 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17966 These options are similar to @option{-fpic} and @option{-fPIC}, but
17967 generated position independent code can be only linked into executables.
17968 Usually these options are used when @option{-pie} GCC option will be
17969 used during linking.
17971 @option{-fpie} and @option{-fPIE} both define the macros
17972 @code{__pie__} and @code{__PIE__}. The macros have the value 1
17973 for @option{-fpie} and 2 for @option{-fPIE}.
17975 @item -fno-jump-tables
17976 @opindex fno-jump-tables
17977 Do not use jump tables for switch statements even where it would be
17978 more efficient than other code generation strategies. This option is
17979 of use in conjunction with @option{-fpic} or @option{-fPIC} for
17980 building code which forms part of a dynamic linker and cannot
17981 reference the address of a jump table. On some targets, jump tables
17982 do not require a GOT and this option is not needed.
17984 @item -ffixed-@var{reg}
17986 Treat the register named @var{reg} as a fixed register; generated code
17987 should never refer to it (except perhaps as a stack pointer, frame
17988 pointer or in some other fixed role).
17990 @var{reg} must be the name of a register. The register names accepted
17991 are machine-specific and are defined in the @code{REGISTER_NAMES}
17992 macro in the machine description macro file.
17994 This flag does not have a negative form, because it specifies a
17997 @item -fcall-used-@var{reg}
17998 @opindex fcall-used
17999 Treat the register named @var{reg} as an allocable register that is
18000 clobbered by function calls. It may be allocated for temporaries or
18001 variables that do not live across a call. Functions compiled this way
18002 will not save and restore the register @var{reg}.
18004 It is an error to used this flag with the frame pointer or stack pointer.
18005 Use of this flag for other registers that have fixed pervasive roles in
18006 the machine's execution model will produce disastrous results.
18008 This flag does not have a negative form, because it specifies a
18011 @item -fcall-saved-@var{reg}
18012 @opindex fcall-saved
18013 Treat the register named @var{reg} as an allocable register saved by
18014 functions. It may be allocated even for temporaries or variables that
18015 live across a call. Functions compiled this way will save and restore
18016 the register @var{reg} if they use it.
18018 It is an error to used this flag with the frame pointer or stack pointer.
18019 Use of this flag for other registers that have fixed pervasive roles in
18020 the machine's execution model will produce disastrous results.
18022 A different sort of disaster will result from the use of this flag for
18023 a register in which function values may be returned.
18025 This flag does not have a negative form, because it specifies a
18028 @item -fpack-struct[=@var{n}]
18029 @opindex fpack-struct
18030 Without a value specified, pack all structure members together without
18031 holes. When a value is specified (which must be a small power of two), pack
18032 structure members according to this value, representing the maximum
18033 alignment (that is, objects with default alignment requirements larger than
18034 this will be output potentially unaligned at the next fitting location.
18036 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
18037 code that is not binary compatible with code generated without that switch.
18038 Additionally, it makes the code suboptimal.
18039 Use it to conform to a non-default application binary interface.
18041 @item -finstrument-functions
18042 @opindex finstrument-functions
18043 Generate instrumentation calls for entry and exit to functions. Just
18044 after function entry and just before function exit, the following
18045 profiling functions will be called with the address of the current
18046 function and its call site. (On some platforms,
18047 @code{__builtin_return_address} does not work beyond the current
18048 function, so the call site information may not be available to the
18049 profiling functions otherwise.)
18052 void __cyg_profile_func_enter (void *this_fn,
18054 void __cyg_profile_func_exit (void *this_fn,
18058 The first argument is the address of the start of the current function,
18059 which may be looked up exactly in the symbol table.
18061 This instrumentation is also done for functions expanded inline in other
18062 functions. The profiling calls will indicate where, conceptually, the
18063 inline function is entered and exited. This means that addressable
18064 versions of such functions must be available. If all your uses of a
18065 function are expanded inline, this may mean an additional expansion of
18066 code size. If you use @samp{extern inline} in your C code, an
18067 addressable version of such functions must be provided. (This is
18068 normally the case anyways, but if you get lucky and the optimizer always
18069 expands the functions inline, you might have gotten away without
18070 providing static copies.)
18072 A function may be given the attribute @code{no_instrument_function}, in
18073 which case this instrumentation will not be done. This can be used, for
18074 example, for the profiling functions listed above, high-priority
18075 interrupt routines, and any functions from which the profiling functions
18076 cannot safely be called (perhaps signal handlers, if the profiling
18077 routines generate output or allocate memory).
18079 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
18080 @opindex finstrument-functions-exclude-file-list
18082 Set the list of functions that are excluded from instrumentation (see
18083 the description of @code{-finstrument-functions}). If the file that
18084 contains a function definition matches with one of @var{file}, then
18085 that function is not instrumented. The match is done on substrings:
18086 if the @var{file} parameter is a substring of the file name, it is
18087 considered to be a match.
18092 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
18096 will exclude any inline function defined in files whose pathnames
18097 contain @code{/bits/stl} or @code{include/sys}.
18099 If, for some reason, you want to include letter @code{','} in one of
18100 @var{sym}, write @code{'\,'}. For example,
18101 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
18102 (note the single quote surrounding the option).
18104 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
18105 @opindex finstrument-functions-exclude-function-list
18107 This is similar to @code{-finstrument-functions-exclude-file-list},
18108 but this option sets the list of function names to be excluded from
18109 instrumentation. The function name to be matched is its user-visible
18110 name, such as @code{vector<int> blah(const vector<int> &)}, not the
18111 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
18112 match is done on substrings: if the @var{sym} parameter is a substring
18113 of the function name, it is considered to be a match. For C99 and C++
18114 extended identifiers, the function name must be given in UTF-8, not
18115 using universal character names.
18117 @item -fstack-check
18118 @opindex fstack-check
18119 Generate code to verify that you do not go beyond the boundary of the
18120 stack. You should specify this flag if you are running in an
18121 environment with multiple threads, but only rarely need to specify it in
18122 a single-threaded environment since stack overflow is automatically
18123 detected on nearly all systems if there is only one stack.
18125 Note that this switch does not actually cause checking to be done; the
18126 operating system or the language runtime must do that. The switch causes
18127 generation of code to ensure that they see the stack being extended.
18129 You can additionally specify a string parameter: @code{no} means no
18130 checking, @code{generic} means force the use of old-style checking,
18131 @code{specific} means use the best checking method and is equivalent
18132 to bare @option{-fstack-check}.
18134 Old-style checking is a generic mechanism that requires no specific
18135 target support in the compiler but comes with the following drawbacks:
18139 Modified allocation strategy for large objects: they will always be
18140 allocated dynamically if their size exceeds a fixed threshold.
18143 Fixed limit on the size of the static frame of functions: when it is
18144 topped by a particular function, stack checking is not reliable and
18145 a warning is issued by the compiler.
18148 Inefficiency: because of both the modified allocation strategy and the
18149 generic implementation, the performances of the code are hampered.
18152 Note that old-style stack checking is also the fallback method for
18153 @code{specific} if no target support has been added in the compiler.
18155 @item -fstack-limit-register=@var{reg}
18156 @itemx -fstack-limit-symbol=@var{sym}
18157 @itemx -fno-stack-limit
18158 @opindex fstack-limit-register
18159 @opindex fstack-limit-symbol
18160 @opindex fno-stack-limit
18161 Generate code to ensure that the stack does not grow beyond a certain value,
18162 either the value of a register or the address of a symbol. If the stack
18163 would grow beyond the value, a signal is raised. For most targets,
18164 the signal is raised before the stack overruns the boundary, so
18165 it is possible to catch the signal without taking special precautions.
18167 For instance, if the stack starts at absolute address @samp{0x80000000}
18168 and grows downwards, you can use the flags
18169 @option{-fstack-limit-symbol=__stack_limit} and
18170 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
18171 of 128KB@. Note that this may only work with the GNU linker.
18173 @item -fsplit-stack
18174 @opindex fsplit-stack
18175 Generate code to automatically split the stack before it overflows.
18176 The resulting program has a discontiguous stack which can only
18177 overflow if the program is unable to allocate any more memory. This
18178 is most useful when running threaded programs, as it is no longer
18179 necessary to calculate a good stack size to use for each thread. This
18180 is currently only implemented for the i386 and x86_64 backends running
18183 When code compiled with @option{-fsplit-stack} calls code compiled
18184 without @option{-fsplit-stack}, there may not be much stack space
18185 available for the latter code to run. If compiling all code,
18186 including library code, with @option{-fsplit-stack} is not an option,
18187 then the linker can fix up these calls so that the code compiled
18188 without @option{-fsplit-stack} always has a large stack. Support for
18189 this is implemented in the gold linker in GNU binutils release 2.21
18192 @item -fleading-underscore
18193 @opindex fleading-underscore
18194 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
18195 change the way C symbols are represented in the object file. One use
18196 is to help link with legacy assembly code.
18198 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
18199 generate code that is not binary compatible with code generated without that
18200 switch. Use it to conform to a non-default application binary interface.
18201 Not all targets provide complete support for this switch.
18203 @item -ftls-model=@var{model}
18204 @opindex ftls-model
18205 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
18206 The @var{model} argument should be one of @code{global-dynamic},
18207 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
18209 The default without @option{-fpic} is @code{initial-exec}; with
18210 @option{-fpic} the default is @code{global-dynamic}.
18212 @item -fvisibility=@var{default|internal|hidden|protected}
18213 @opindex fvisibility
18214 Set the default ELF image symbol visibility to the specified option---all
18215 symbols will be marked with this unless overridden within the code.
18216 Using this feature can very substantially improve linking and
18217 load times of shared object libraries, produce more optimized
18218 code, provide near-perfect API export and prevent symbol clashes.
18219 It is @strong{strongly} recommended that you use this in any shared objects
18222 Despite the nomenclature, @code{default} always means public; i.e.,
18223 available to be linked against from outside the shared object.
18224 @code{protected} and @code{internal} are pretty useless in real-world
18225 usage so the only other commonly used option will be @code{hidden}.
18226 The default if @option{-fvisibility} isn't specified is
18227 @code{default}, i.e., make every
18228 symbol public---this causes the same behavior as previous versions of
18231 A good explanation of the benefits offered by ensuring ELF
18232 symbols have the correct visibility is given by ``How To Write
18233 Shared Libraries'' by Ulrich Drepper (which can be found at
18234 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
18235 solution made possible by this option to marking things hidden when
18236 the default is public is to make the default hidden and mark things
18237 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
18238 and @code{__attribute__ ((visibility("default")))} instead of
18239 @code{__declspec(dllexport)} you get almost identical semantics with
18240 identical syntax. This is a great boon to those working with
18241 cross-platform projects.
18243 For those adding visibility support to existing code, you may find
18244 @samp{#pragma GCC visibility} of use. This works by you enclosing
18245 the declarations you wish to set visibility for with (for example)
18246 @samp{#pragma GCC visibility push(hidden)} and
18247 @samp{#pragma GCC visibility pop}.
18248 Bear in mind that symbol visibility should be viewed @strong{as
18249 part of the API interface contract} and thus all new code should
18250 always specify visibility when it is not the default; i.e., declarations
18251 only for use within the local DSO should @strong{always} be marked explicitly
18252 as hidden as so to avoid PLT indirection overheads---making this
18253 abundantly clear also aids readability and self-documentation of the code.
18254 Note that due to ISO C++ specification requirements, operator new and
18255 operator delete must always be of default visibility.
18257 Be aware that headers from outside your project, in particular system
18258 headers and headers from any other library you use, may not be
18259 expecting to be compiled with visibility other than the default. You
18260 may need to explicitly say @samp{#pragma GCC visibility push(default)}
18261 before including any such headers.
18263 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
18264 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
18265 no modifications. However, this means that calls to @samp{extern}
18266 functions with no explicit visibility will use the PLT, so it is more
18267 effective to use @samp{__attribute ((visibility))} and/or
18268 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
18269 declarations should be treated as hidden.
18271 Note that @samp{-fvisibility} does affect C++ vague linkage
18272 entities. This means that, for instance, an exception class that will
18273 be thrown between DSOs must be explicitly marked with default
18274 visibility so that the @samp{type_info} nodes will be unified between
18277 An overview of these techniques, their benefits and how to use them
18278 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
18280 @item -fstrict-volatile-bitfields
18281 @opindex fstrict-volatile-bitfields
18282 This option should be used if accesses to volatile bitfields (or other
18283 structure fields, although the compiler usually honors those types
18284 anyway) should use a single access of the width of the
18285 field's type, aligned to a natural alignment if possible. For
18286 example, targets with memory-mapped peripheral registers might require
18287 all such accesses to be 16 bits wide; with this flag the user could
18288 declare all peripheral bitfields as ``unsigned short'' (assuming short
18289 is 16 bits on these targets) to force GCC to use 16 bit accesses
18290 instead of, perhaps, a more efficient 32 bit access.
18292 If this option is disabled, the compiler will use the most efficient
18293 instruction. In the previous example, that might be a 32-bit load
18294 instruction, even though that will access bytes that do not contain
18295 any portion of the bitfield, or memory-mapped registers unrelated to
18296 the one being updated.
18298 If the target requires strict alignment, and honoring the field
18299 type would require violating this alignment, a warning is issued.
18300 If the field has @code{packed} attribute, the access is done without
18301 honoring the field type. If the field doesn't have @code{packed}
18302 attribute, the access is done honoring the field type. In both cases,
18303 GCC assumes that the user knows something about the target hardware
18304 that it is unaware of.
18306 The default value of this option is determined by the application binary
18307 interface for the target processor.
18313 @node Environment Variables
18314 @section Environment Variables Affecting GCC
18315 @cindex environment variables
18317 @c man begin ENVIRONMENT
18318 This section describes several environment variables that affect how GCC
18319 operates. Some of them work by specifying directories or prefixes to use
18320 when searching for various kinds of files. Some are used to specify other
18321 aspects of the compilation environment.
18323 Note that you can also specify places to search using options such as
18324 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
18325 take precedence over places specified using environment variables, which
18326 in turn take precedence over those specified by the configuration of GCC@.
18327 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
18328 GNU Compiler Collection (GCC) Internals}.
18333 @c @itemx LC_COLLATE
18335 @c @itemx LC_MONETARY
18336 @c @itemx LC_NUMERIC
18341 @c @findex LC_COLLATE
18342 @findex LC_MESSAGES
18343 @c @findex LC_MONETARY
18344 @c @findex LC_NUMERIC
18348 These environment variables control the way that GCC uses
18349 localization information that allow GCC to work with different
18350 national conventions. GCC inspects the locale categories
18351 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
18352 so. These locale categories can be set to any value supported by your
18353 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
18354 Kingdom encoded in UTF-8.
18356 The @env{LC_CTYPE} environment variable specifies character
18357 classification. GCC uses it to determine the character boundaries in
18358 a string; this is needed for some multibyte encodings that contain quote
18359 and escape characters that would otherwise be interpreted as a string
18362 The @env{LC_MESSAGES} environment variable specifies the language to
18363 use in diagnostic messages.
18365 If the @env{LC_ALL} environment variable is set, it overrides the value
18366 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
18367 and @env{LC_MESSAGES} default to the value of the @env{LANG}
18368 environment variable. If none of these variables are set, GCC
18369 defaults to traditional C English behavior.
18373 If @env{TMPDIR} is set, it specifies the directory to use for temporary
18374 files. GCC uses temporary files to hold the output of one stage of
18375 compilation which is to be used as input to the next stage: for example,
18376 the output of the preprocessor, which is the input to the compiler
18379 @item GCC_EXEC_PREFIX
18380 @findex GCC_EXEC_PREFIX
18381 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
18382 names of the subprograms executed by the compiler. No slash is added
18383 when this prefix is combined with the name of a subprogram, but you can
18384 specify a prefix that ends with a slash if you wish.
18386 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
18387 an appropriate prefix to use based on the pathname it was invoked with.
18389 If GCC cannot find the subprogram using the specified prefix, it
18390 tries looking in the usual places for the subprogram.
18392 The default value of @env{GCC_EXEC_PREFIX} is
18393 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
18394 the installed compiler. In many cases @var{prefix} is the value
18395 of @code{prefix} when you ran the @file{configure} script.
18397 Other prefixes specified with @option{-B} take precedence over this prefix.
18399 This prefix is also used for finding files such as @file{crt0.o} that are
18402 In addition, the prefix is used in an unusual way in finding the
18403 directories to search for header files. For each of the standard
18404 directories whose name normally begins with @samp{/usr/local/lib/gcc}
18405 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
18406 replacing that beginning with the specified prefix to produce an
18407 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
18408 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
18409 These alternate directories are searched first; the standard directories
18410 come next. If a standard directory begins with the configured
18411 @var{prefix} then the value of @var{prefix} is replaced by
18412 @env{GCC_EXEC_PREFIX} when looking for header files.
18414 @item COMPILER_PATH
18415 @findex COMPILER_PATH
18416 The value of @env{COMPILER_PATH} is a colon-separated list of
18417 directories, much like @env{PATH}. GCC tries the directories thus
18418 specified when searching for subprograms, if it can't find the
18419 subprograms using @env{GCC_EXEC_PREFIX}.
18422 @findex LIBRARY_PATH
18423 The value of @env{LIBRARY_PATH} is a colon-separated list of
18424 directories, much like @env{PATH}. When configured as a native compiler,
18425 GCC tries the directories thus specified when searching for special
18426 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
18427 using GCC also uses these directories when searching for ordinary
18428 libraries for the @option{-l} option (but directories specified with
18429 @option{-L} come first).
18433 @cindex locale definition
18434 This variable is used to pass locale information to the compiler. One way in
18435 which this information is used is to determine the character set to be used
18436 when character literals, string literals and comments are parsed in C and C++.
18437 When the compiler is configured to allow multibyte characters,
18438 the following values for @env{LANG} are recognized:
18442 Recognize JIS characters.
18444 Recognize SJIS characters.
18446 Recognize EUCJP characters.
18449 If @env{LANG} is not defined, or if it has some other value, then the
18450 compiler will use mblen and mbtowc as defined by the default locale to
18451 recognize and translate multibyte characters.
18455 Some additional environments variables affect the behavior of the
18458 @include cppenv.texi
18462 @node Precompiled Headers
18463 @section Using Precompiled Headers
18464 @cindex precompiled headers
18465 @cindex speed of compilation
18467 Often large projects have many header files that are included in every
18468 source file. The time the compiler takes to process these header files
18469 over and over again can account for nearly all of the time required to
18470 build the project. To make builds faster, GCC allows users to
18471 `precompile' a header file; then, if builds can use the precompiled
18472 header file they will be much faster.
18474 To create a precompiled header file, simply compile it as you would any
18475 other file, if necessary using the @option{-x} option to make the driver
18476 treat it as a C or C++ header file. You will probably want to use a
18477 tool like @command{make} to keep the precompiled header up-to-date when
18478 the headers it contains change.
18480 A precompiled header file will be searched for when @code{#include} is
18481 seen in the compilation. As it searches for the included file
18482 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
18483 compiler looks for a precompiled header in each directory just before it
18484 looks for the include file in that directory. The name searched for is
18485 the name specified in the @code{#include} with @samp{.gch} appended. If
18486 the precompiled header file can't be used, it is ignored.
18488 For instance, if you have @code{#include "all.h"}, and you have
18489 @file{all.h.gch} in the same directory as @file{all.h}, then the
18490 precompiled header file will be used if possible, and the original
18491 header will be used otherwise.
18493 Alternatively, you might decide to put the precompiled header file in a
18494 directory and use @option{-I} to ensure that directory is searched
18495 before (or instead of) the directory containing the original header.
18496 Then, if you want to check that the precompiled header file is always
18497 used, you can put a file of the same name as the original header in this
18498 directory containing an @code{#error} command.
18500 This also works with @option{-include}. So yet another way to use
18501 precompiled headers, good for projects not designed with precompiled
18502 header files in mind, is to simply take most of the header files used by
18503 a project, include them from another header file, precompile that header
18504 file, and @option{-include} the precompiled header. If the header files
18505 have guards against multiple inclusion, they will be skipped because
18506 they've already been included (in the precompiled header).
18508 If you need to precompile the same header file for different
18509 languages, targets, or compiler options, you can instead make a
18510 @emph{directory} named like @file{all.h.gch}, and put each precompiled
18511 header in the directory, perhaps using @option{-o}. It doesn't matter
18512 what you call the files in the directory, every precompiled header in
18513 the directory will be considered. The first precompiled header
18514 encountered in the directory that is valid for this compilation will
18515 be used; they're searched in no particular order.
18517 There are many other possibilities, limited only by your imagination,
18518 good sense, and the constraints of your build system.
18520 A precompiled header file can be used only when these conditions apply:
18524 Only one precompiled header can be used in a particular compilation.
18527 A precompiled header can't be used once the first C token is seen. You
18528 can have preprocessor directives before a precompiled header; you can
18529 even include a precompiled header from inside another header, so long as
18530 there are no C tokens before the @code{#include}.
18533 The precompiled header file must be produced for the same language as
18534 the current compilation. You can't use a C precompiled header for a C++
18538 The precompiled header file must have been produced by the same compiler
18539 binary as the current compilation is using.
18542 Any macros defined before the precompiled header is included must
18543 either be defined in the same way as when the precompiled header was
18544 generated, or must not affect the precompiled header, which usually
18545 means that they don't appear in the precompiled header at all.
18547 The @option{-D} option is one way to define a macro before a
18548 precompiled header is included; using a @code{#define} can also do it.
18549 There are also some options that define macros implicitly, like
18550 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
18553 @item If debugging information is output when using the precompiled
18554 header, using @option{-g} or similar, the same kind of debugging information
18555 must have been output when building the precompiled header. However,
18556 a precompiled header built using @option{-g} can be used in a compilation
18557 when no debugging information is being output.
18559 @item The same @option{-m} options must generally be used when building
18560 and using the precompiled header. @xref{Submodel Options},
18561 for any cases where this rule is relaxed.
18563 @item Each of the following options must be the same when building and using
18564 the precompiled header:
18566 @gccoptlist{-fexceptions}
18569 Some other command-line options starting with @option{-f},
18570 @option{-p}, or @option{-O} must be defined in the same way as when
18571 the precompiled header was generated. At present, it's not clear
18572 which options are safe to change and which are not; the safest choice
18573 is to use exactly the same options when generating and using the
18574 precompiled header. The following are known to be safe:
18576 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
18577 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
18578 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
18583 For all of these except the last, the compiler will automatically
18584 ignore the precompiled header if the conditions aren't met. If you
18585 find an option combination that doesn't work and doesn't cause the
18586 precompiled header to be ignored, please consider filing a bug report,
18589 If you do use differing options when generating and using the
18590 precompiled header, the actual behavior will be a mixture of the
18591 behavior for the options. For instance, if you use @option{-g} to
18592 generate the precompiled header but not when using it, you may or may
18593 not get debugging information for routines in the precompiled header.